A Descriptive Survey Research Study of The Student Characteristi
A Descriptive Survey Research Study of The Student Characteristi
A Descriptive Survey Research Study of The Student Characteristi
UKnowledge
Theses and Dissertations--Science, Technology,
Science, Technology, Engineering, and Mathematics
Engineering, and Mathematics (STEM) Education
(STEM) Education
2012
Recommended Citation
Locklear, Tonja Motley, "A DESCRIPTIVE, SURVEY RESEARCH STUDY OF THE STUDENT CHARACTERISTICS
INFLUENCING THE FOUR THEORETICAL SOURCES OF MATHEMATICAL SELF-EFFICACY OF COLLEGE
FRESHMEN" (2012). Theses and Dissertations--Science, Technology, Engineering, and Mathematics (STEM) Education. Paper 1.
http://uknowledge.uky.edu/stem_etds/1
This Doctoral Dissertation is brought to you for free and open access by the Science, Technology, Engineering, and Mathematics (STEM) Education at
UKnowledge. It has been accepted for inclusion in Theses and Dissertations--Science, Technology, Engineering, and Mathematics (STEM) Education
by an authorized administrator of UKnowledge. For more information, please contact [email protected].
STUDENT AGREEMENT:
I represent that my thesis or dissertation and abstract are my original work. Proper attribution has been
given to all outside sources. I understand that I am solely responsible for obtaining any needed copyright
permissions. I have obtained and attached hereto needed written permission statements(s) from the
owner(s) of each third-party copyrighted matter to be included in my work, allowing electronic
distribution (if such use is not permitted by the fair use doctrine).
I hereby grant to The University of Kentucky and its agents the non-exclusive license to archive and make
accessible my work in whole or in part in all forms of media, now or hereafter known. I agree that the
document mentioned above may be made available immediately for worldwide access unless a
preapproved embargo applies.
I retain all other ownership rights to the copyright of my work. I also retain the right to use in future
works (such as articles or books) all or part of my work. I understand that I am free to register the
copyright to my work.
REVIEW, APPROVAL AND ACCEPTANCE
The document mentioned above has been reviewed and accepted by the students advisor, on behalf of
the advisory committee, and by the Director of Graduate Studies (DGS), on behalf of the program; we
verify that this is the final, approved version of the students dissertation including all changes required
by the advisory committee. The undersigned agree to abide by the statements above.
Tonja Motley Locklear, Student
Dr. Margaret J. Mohr-Schroeder, Major Professor
Dr. Robert Shapiro, Director of Graduate Studies
____________________________________
DISSERTATION
_____________________________________
ABSTRACT OF DISSERTATION
BY
Tonja Motley Locklear
I would like to dedicate this work to my wonderful children, Hannah and Peyton,
who had to suffer through my frustrations, lack of quality time,
and stressful deadlines. May you both realize that nothing is impossible
as long as you have the support of loving family and friends,
set your goals, work hard and pray even harder.
I LOVE YOU!
ACKNOWLEDGEMENTS
Even though this paper was my own work, it would not have been possible without the
support and guidance of many people:
ACCLAIM Doctoral Program Coordinators and Professors Without this particular
program offered by the collaboration of many outstanding professors from five
universities, I would NEVER have had the opportunity to fulfill my dream. THANK
YOU for your innovation and dedication to education!
ACCLAIM Cohort III Without your own stories, experiences, knowledge and support,
my understanding of mathematics teacher education would not have evolved. THANK
YOU for making me a stronger mathematics educator!
Study Buddies Without your support and encouragement throughout each of the 5-week
summer sessions, as well as all of the online courses, I would not have felt as confident to
press forward. Each of you has exemplified courage, strength and friendship through your
own personal obstacles and setbacks. THANK YOU for always being there to listen!
Dr. Margaret Schroeder and Dr. Carl Lee Without your guidance and support as
ACCLAIM professors and co-chairs of my dissertation committee, I would not have felt
as prepared and confident during my defense. I hope you will forgive my tireless need to
press forward, which may have rocked the boat on occasion. THANK YOU for your
support and encouragement throughout this strenuous process!
Dr. Kenneth Tyler and Dr. Jennifer Wilhelm Without your additional guidance and
knowledge, I would not have focused on generating such a comprehensive study. Your
questions and discussions during my qualifying examination and my proposal defense
created additional layers of my knowledge of self-efficacy and my study of its sources.
THANK YOU for illuminating the areas that needed more investigation!
Mom and Dad Without your support, encouragement, and love, I would not have been
able to accomplish this goal. Knowing that I could always count on you to watch the
children provided such a peace of mind that allowed me to focus on the work at hand.
THANK YOU for providing the foundation during my childhood that said that hard work
will allow you to accomplish your goals!
My Husband Without your understanding and support during this past year, I would not
have been able to complete my dissertation. You allowed me to spend COUNTLESS
hours working on this little paper without MUCH complaining during our FIRST year
of marriage. Your support and love has meant so much to me. THANK YOU for
allowing me the space to complete my professional dream!!!!
God THANK YOU for the wisdom, the strength, the perseverance and the ability!
iii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ............................................................................................... iii
LIST OF TABLES ............................................................................................................. vi
LIST OF FIGURES .......................................................................................................... vii
Chapter I: Introduction ........................................................................................................ 1
Theoretical Framework ................................................................................................... 2
Self-Efficacy ................................................................................................................ 2
Sources of Self-Efficacy .............................................................................................. 3
Statement of the Problem ................................................................................................ 4
Purpose of the Study ....................................................................................................... 5
Research Question ........................................................................................................... 5
Significance of the Study ................................................................................................ 6
Definition of Terms ......................................................................................................... 6
Assumptions .................................................................................................................... 7
Limitations ...................................................................................................................... 7
Organization of the Study ............................................................................................... 8
Chapter II: Review of Literature ......................................................................................... 9
Mathematical Self-Efficacy............................................................................................. 9
Sources of Mathematical Self-Efficacy......................................................................... 13
Mastery experience .................................................................................................... 14
Vicarious experience ................................................................................................. 16
Verbal and social persuasion ..................................................................................... 19
Physiological and emotional states ............................................................................ 20
Conclusion ..................................................................................................................... 20
Chapter III: Methodology ................................................................................................. 23
Research Design ............................................................................................................ 23
Population and sample............................................................................................... 23
Instrumentation .......................................................................................................... 25
Pilot study .................................................................................................................. 28
Procedures ................................................................................................................. 28
Data Analysis ................................................................................................................ 30
Measures .................................................................................................................... 30
Statistical analysis ..................................................................................................... 31
Validity .......................................................................................................................... 33
Chapter IV: Results and Analysis ..................................................................................... 35
Missing Data ................................................................................................................. 35
Demographics ................................................................................................................ 36
Data Screening .............................................................................................................. 36
Confirmatory Factor Analysis ....................................................................................... 41
iv
LIST OF TABLES
Table 4.1 Descriptive statistics for the seven dependent variables ..................................44
Table 4.2 2x2x3 Factorial ANOVA on Mastery Experience with student
characteristics as independent variables ................................................................48
Table 4.3 3x4x4 Factorial ANOVA on Mastery Experience with performance
variables as the independent variables ...................................................................49
Table 4.4 2x2x3 Factorial ANOVA on the dependent variable, vicarious experience,
with student characteristics as independent variables ............................................51
Table 4.5 3x4x4 Factorial ANOVA on the dependent variable, vicarious experience,
with performance variables as the independent variables......................................52
Table 4.6 2x2x3 Factorial ANOVAs on the dependent variable, verbal and social
persuasions, with the student characteristics as the independent variables ...........53
Table 4.7 3x4x4 Factorial ANOVA on the dependent variable, verbal and social
persuasions, with the performance variables as the independent variables ...........54
Table 4.8 2x2x3 Factorial ANOVA on the dependent variable, physiological and
emotional states, with the student characteristics as the independent variable......55
Table 4.9 3x4x4 Factorial ANOVA on the dependent variable, physiological and
emotional states, with the performance variables as the independent variables ....56
Table 4.10 2x2x3 Factorial ANOVA on the dependent variable, vicarious experience
adult role models, with the student characteristics as the independent
variables .................................................................................................................57
Table 4.11 3x4x4 Factorial ANOVA with the dependent variable, vicarious
experience adult role models, with the performance variables as the
independent variables.............................................................................................57
Table 4.12 2x2x3 Factorial ANOVA with dependent variable, vicarious experience
peer role model, with student characteristics as the independent variables ........58
Table 4.13 3x4x4 Factorial ANOVA with the dependent variable, vicarious
experience peer role model, with performance variables as the independent
variables .................................................................................................................58
Table 4.14 One-Way ANOVA on the dependent variable, vicarious experience self
role model, with the independent variable, gender ................................................60
Table 4.15 One-Way ANOVA on the dependent variable, vicarious experience self
role model, with the independent variable, race ....................................................60
Table 4.16 One-Way ANOVA on the dependent variable, vicarious experience self
role model, with the independent variable, hometown environment .....................60
Table 4.17 3x4x4 Factorial ANOVA with the dependent variable, vicarious
experience with self role model, with the performance variables as the
independent variables.............................................................................................61
vi
LIST OF FIGURES
Figure 4.1 Sexual Orientation Demographics ................................................................... 36
Figure 4.2 High School GPA Demographics (HSGPA) ................................................... 37
Figure 4.3 College GPA Demographics (COLGPA) ........................................................ 37
Figure 4.4 High School GPA Demographics Reorganized (HSGPA2) ............................ 38
Figure 4.5 College GPA Demographics Reorganized (COLGPA2) ................................ 38
Figure 4.6 Letter grade received in math course taken the previous semester (classgrd) 39
Figure 4.7 Confirmatory Factor Analysis Model for Maximum Likelihood and Bayesian
estimation of the twenty four survey questions ................................................. 42
vii
Chapter I: Introduction
Since the emergence of self-efficacy as a component of Albert Banduras social
cognitive theory in 1977, much research has been conducted on how this construct affects
learning. Similarly, with the emergence over the past 30 years of mathematics being a
significant subject of discussion in education and politics as seen by publications such as
Nation at Risk, NCTMs Principles and Standards for School Mathematics, No Child Left
Behind, and more recently the, Common Core State Standards Initiative much research
has been directed towards understanding and improving the mathematical achievement of
United States students. It is no surprise then that these two areas of research have
resulted in many studies focusing on the effects of self-efficacy on mathematics
achievement. Consequently, the research has shown in school mathematics that
perceived self-efficacy contributes to academic performance irrespective of the level of
intellectual ability, and correlates strongly with academic outcomes, such as performance
in problem solving, attitudes towards mathematics and math anxiety (Michaelides, 2008,
p. 222).
Even though studying the effects of self-efficacy on mathematics achievement of
students in elementary, middle and even high school could result in interventions that
may produce improvements for not only the students but also mathematics education in
general, studying the effects on the collegiate level is just as significant and important.
The number of students enrolling in colleges or universities over the past few decades has
steadily increased and students academic choices regarding mathematics have been
acknowledged to not only affect a students choice in a college major, but also to
influence a students likelihood for completing his or her college education (Hall &
Ponton, 2005). If the issues regarding self-efficacy and mathematics achievement are not
resolved during the K-12 years, then those same issues will be found on the collegiate
level and may become worse by the increase in stress associated with freshmen students.
In Pajares (1996) review of the educational research of academic self-efficacy, he
acknowledged the sound connection between self-efficacy and academic performances
and achievement established through abundant research. Since this connection has been
well established, the focus needs to shift toward research on how to enhance students
level of self-efficacy through various sources of information. Lent, Lopez, Brown and
Gore (1996) determined through confirmatory factor analysis that the four theorized
sources of mathematical self-efficacy represent different types of information and,
therefore, can be analyzed separately. Additionally, they posited the need to clarify how
the efficacy sources are structured in other populations and performance domains and
whether the theoretical distinctions among the sources can inform interventions aimed at
modifying self-efficacy percepts (Lent et al., 1996, p. 306).
Theoretical Framework
Self-Efficacy. Self-efficacy is a component of Albert Banduras social cognitive
theory and is referred to as the beliefs in ones capabilities to organize and execute the
courses of action required to produce given attainments (Bandura, 1997, p. 3). Thus, the
mathematical self-efficacy of students is their belief about their capabilities to control the
outcome regarding their mathematical performance. For this research study the definition
of mathematical self-efficacy will be the level of an individuals belief in his/her
competence to attain a favorable outcome regarding their mathematical performance.
Bandura goes on to state,
into three subsections. Verbal and social persuasions refers to the cognitive processing of
the faith, or lack thereof, others have in an individuals abilities. The knowledge and
credibility of the other person makes a difference in the level of influence of this source
of self-efficacy. Finally, physiological and emotional states refers to the cognitive
processing of an individuals capabilities based on somatic information. Anxiety is a
common somatic indicator associated with mathematics, which may affect how
efficacious a person feels toward mathematics success (Bandura, 1997).
Statement of the Problem
Post-secondary education faces multiple challenges in providing the best
environment to assist students from matriculation to graduation. Mathematics has been
one area of contention with college students for years. Knowing how to build a sense of
efficacy and how it works provides further guidelines for structuring experiences that
enable people to realize desired personal and social changes (Bandura, 2006, p. 319). In
order to create the experiences to enhance the mathematical self-efficacy of college
students, more information is needed on how the different sources of self-efficacy impact
various people and/or groups of people.
Bandura (1997) posited that mastery experience is the strongest source of selfefficacy, but Stevens et al. (2004), stated that previous mathematics achievement
(mastery experience) influenced the self-efficacy of Hispanic students in their study, but
did not have a statistically significant impact on the Caucasian students. However, they
used the students self-reported prior mathematics grades to represent the mastery
experience construct, which does not evaluate the students interpretation of those grades.
Their study and many others (Matsui, Matsui, & Ohnishi, 1990; Ozyurek, 2005) have
achievement score may not speak to the actual ability of the students. Additionally, they
reported that the self-efficacy of students taking online classes was not substantially
different than from students who took a traditionally delivered course. However, the selfefficacy survey questions were only given to students who took a course online and the
survey described was not a known, calibrated assessment of self-efficacy. Thus, this
generalized statement has no merit within the confines of the research study.
Spence & Usher (2007) attempted to determine how motivation variables,
mathematics self-efficacy and a particular computer courseware (MyMathLab) affected
student achievement. Their study included 164 students who were enrolled in either one
of the eight sections of traditionally taught courses or one of the eight sections of online
courses at a particular university and who completed both surveys and took the final
exam. The final exam was the assessment tool for determining mathematics achievement,
because all students were required to take this 40-item multiple-choice departmental test.
Even though Spence and Usher acknowledged the use of Banduras Guide for
Constructing Self-Efficacy Scales as their reasoning for the questions used within the
surveys to assess self-efficacy, the survey was not a known, calibrated measure.
Spence and Usher (2007) found that the mathematics self-efficacy was higher in
those enrolled in the traditional courses than those in the online courses. The students in
the traditional courses also scored higher on their final exams. When researchers
controlled for mathematics self-efficacy, they claimed that the poorer performance of the
online students was associated with their lower self-efficacy beliefs. Compared to the
other variables (self-efficacy for self-regulation, computer self-efficacy, computer
playfulness, engagement and age), they determined that mathematics self-efficacy was
10
the strongest predictor of achievement. However, there were some validity issues that
need to be considered within the context of this study. Those students taking courses
online were required to take a written mathematics test. This change in assessment style
from computerized to written may account for the poorer scores on the final exam.
Additionally, those students enrolled in the traditional courses were allowed to use
MyMathLab courseware as a supplement, which could confound the results of the
achievement assessment.
Hall and Ponton (2005) focused on the comparison of the mathematics selfefficacy of freshmen students taking developmental mathematics courses with those
taking Calculus I. After hypothesizing that participation in the study would have no
positive or negative effect on the students grade for the semester, 80 students within the
four Calculus I sections and 105 students within the four developmental sections agreed
to participate. The students were given two subscales (Mathematics Tasks subscale and
Mathematics Courses subscale) of the revised Mathematical Self-Efficacy Scale (MSES),
which produced their overall self-efficacy score. Using an independent t-test and an
ANOVA, the results showed a statistically significant difference between the selfefficacy of freshmen students in the developmental mathematics courses and those in the
Calculus I courses with the latter having the higher level of self-efficacy. Additionally,
gender did not statistically significantly affect students mathematical self-efficacy in
either the developmental or Calculus I courses. Even though the results of this study did
not appear to be confounded by internal validity issues, the small sample size limits the
generalizability of these results.
11
12
mentioned the necessity of measuring self-efficacy specifically and within a close time
frame of the mathematics achievement measure. The authors specifically mention this
limitation but explained that this microanalysis was not suitable for their particular
research study on whether self-beliefs more generally predicted student achievement.
This strategy, coupled with the fact that academic self-beliefs were measured together to
produce the key finding that prior knowledge predicted student achievement over all
other variables, makes this result seem less reliable.
The articles mentioned in this section of the literature review indicate the
necessity of analyzing the mathematical self-efficacy of students based on the tenets
established by Bandura (1997) through a known and calibrated assessment scale. When
attention is not given to the tenets of the social cognitive theory, conflicting results can
arise. Wadsworth, et al. (2007) determined that the mathematical self-efficacy of students
within a traditional or online course was basically the same, but Spence and Usher (2007)
found the mathematical self-efficacy of the students in the traditional course to be higher
than those in the online course. Hailikari, et al. (2008) completely disregarded Banduras
warning of analyzing the construct within a close proximity of the assessment of the
academic achievement variable, which creates a very unreliable conclusion that
mathematical self-efficacy was not as influential as prior knowledge on mathematical
achievement. The following section on the sources of mathematical self-efficacy
continues this analysis based on the tenets of the social cognitive theory.
Sources of Mathematical Self-Efficacy
The four theorized sources of self-efficacy (mastery experience, vicarious
experience, verbal and social persuasion, and physiological and emotional states) have
13
14
Luzzo et al. (1999) conducted their study with four treatment conditions analyzed
with pre- and post-test assessments of mathematics/science self-efficacy. The treatment
groups consisted of participants taking a test of incomplete number series to enhance the
performance or mastery experience source, the participants watching a 15-minute video
of two successful graduates with similar backgrounds and stories to enhance the vicarious
experience source, participants performing a combination of the two, and a control group
with no intervention. The study determined that the performance accomplishment
treatment had a statistically stronger impact on the participants than the videos of adult
models. However, just viewing fifteen minutes of videos of successful strangers may not
constitute the idea of a genuine role model with which to enhance self-efficacy through
vicarious means. In addition to the treatment modes, the assessments occurred three times
(pre-, post-, and immediately after the treatment) and the final two were only four weeks
apart. This posed a test/retest threat to the internal validity of the study.
In an attempt to study the factors that inform the mathematical self-efficacy of
students, Lent, Brown, Gover, & Nijjer (1996) had their participants answer questions on
demographics, career aspirations and mathematics self-efficacy. The mathematics selfefficacy was evaluated by rating their confidence to obtain a B or better in various
mathematics courses; however, this is not in line with the theoretical nature of selfefficacy. Perceived self-efficacy is a judgment of capability to execute given types of
performances; outcome expectations are judgments about the outcomes that are likely to
flow from such performances (Bandura, 2006, p. 309). The study also contained a
qualitative component by asking the participants to do a thought-listing of the reasons for
their confidence ratings. The researchers then categorized the various statements within
15
the four theorized sources, as well as other emerging categories such as interest, effort,
teacher quality, etc. Their analysis consisted of determining the percentage of comments
per category. This study determined that personal performance (mastery experience)
accounted for 58% of the responses. However, when focusing on the specifics of
obtaining a B or better in a mathematics course, it is understandable that the responses
related to confidence would fall more often within the performance category. Even
though the qualitative nature of this study provides data that supports the methodological
basis of the four sources of mathematical self-efficacy and retrieves the data in a manner
that is underutilized within this specific field of the social cognitive theory, the types of
questions in the written section may have lead to skewed responses towards performance
criteria.
Not only has the mastery experience construct been analyzed within research
studies by specific quantitative scales (Lent, Lopez, & Bieschke, 1991; Lent, Lopez,
Brown, & Gore, 1996; Usher & Pajares, Sources of self-efficacy in mathematics: A
validation study, 2009), other studies have used prior grades, successful completion of
number series and a thought-listing analysis to provide evidence of the influence of the
mastery experience source over the other sources. Even though the variety of methods
corroborate Banduras belief that mastery experience is the most influential source of
self-efficacy, researchers must still analyze the construct in a manner consistent with the
tenets of the theory.
Vicarious experience. Vicarious experience has consistently had a low to modest
Cronbach alpha reliability coefficient value within research studies (Hodges & Murphy,
2009; Lent, Lopez, & Bieschke, 1991; Lent, Lopez, Brown, & Gore, 1996; Matsui,
16
Matsui, & Ohnishi, 1990). Usher and Pajares (2009) contend that the lower coefficient
alphas may be caused by research studies focusing on only peer or adult models, but not
both. However, some research studies (e.g., Hodges & Murphy, 2009; Zeldin & Pajares,
2000) found the vicarious experience construct to be most influential.
Even though Zeldin & Pajares (2000) focused their research on 15 women who
were already involved in careers within the STEM (Science, Technology, Engineering
and Mathematics) areas, this qualitative study seemed appropriate to include within this
review. The interviews of the women revealed that their vicarious experiences and their
verbal and social persuasions were highly influential in their success within the STEM
career paths. Since the women were interviewed after establishing careers in the STEM
areas, mastery experience, which is posited by Bandura (1997) to be the most influential
source of information, may not be as relevant to them in retrospect. The grades and the
academic accomplishments may fade over time; whereas, the personal connections and
encouragement may grow stronger.
The sample of women was purposive for this study and had to meet the
requirements selected by the researchers. Each interview consisted of the same nine
questions and was recorded and transcribed verbatim. The researchers explained how
they tried to account for internal validity by having other colleagues with familiarity of
educational issues, but not self-efficacy theory, read through the transcripts and
determine if other themes emerged from the data.
The women in the study noted teachers who had been influential in their academic
pursuits of careers in STEM areas. These women also discussed academic resiliency as it
related to how they continued to pursue their majors in college and their careers upon
17
graduating when obstacles would try to deter them. The purposeful sample and the small
sample size did not allow for generalizability. However, being able to interview these
women after navigating through the academic world and emerging with a career in a
STEM area provided additional insight into how self-efficacy could play a significant
role in not only academic achievement but also within academic and life-long success.
Hodges & Murphy (2009) conducted a study to explore the sources of selfefficacy of students enrolled in a technology-intensive asynchronous college algebra
course. The Sources of Mathematics Self-Efficacy (SMSE) scale developed by Lent et al.
(1991) was used to determine the scores for the four theoretical sources, which were used
as predictor variables within the study. The dependent variable was the Self-Efficacy for
Learning Mathematics Asynchronously survey developed by Hodges (2008). The
regression analysis showed that vicarious experiences and physiological/affective states
were the only two statistically significant predictors. It seems reasonable that how a
student feels throughout the duration of taking an asynchronous course would have a
significant impact on how efficacious he/she is about learning in an asynchronous
environment. Additionally, learning in a non-traditional environment could be influenced
by the successes of other students who have taken similar courses asynchronously. Since
the dependent variable pertains to the self-efficacy of learning mathematics
asynchronously, which usually indicates an online course, and the data was collected
using an online survey, the reliability of the results may be questionable. Students who
feel less efficacious about taking an asynchronous course may not feel efficacious about
taking an online survey.
18
Since the vicarious experience construct has had low to modest Cronbachs alpha
reliability coefficients in research studies , (Hodges & Murphy, 2009; Lent, Lopez, &
Bieschke, 1991; Lent, Lopez, Brown, & Gore, 1996; Matsui, Matsui, & Ohnishi, 1990) it
is important to determine a way of assessing this source of mathematical self-efficacy
that will not only be consistent with the tenets of the theory but also will provide a more
consistent measure. Zeldin and Pajares (2000) were able to obtain results of the
influential impact of the vicarious experience construct on the women in their study
through an interview process. Qualitative studies may provide a more informative
measure of this construct because it does not restrict the role models to only adults, peers,
or teachers. Students are influenced by other role models in society as well, but the
research on the sources of self-efficacy does not usually include those options.
Verbal and social persuasion. Verbal and social persuasion has been studied by
researchers focused on encouragement by teachers, parents and friends (Hodges &
Murphy, 2009; Lent, Lopez, Brown, & Gore, 1996; Matsui, Matsui, & Ohnishi, 1990).
Encouraging statements were extremely evident in the stories told by the women with
careers in STEM areas in the qualitative study by Zeldin & Pajares (2000). The
encouragement was described with regards to not only mathematics but also a confidence
that others believed they could succeed in anything. Those encouraging words
strengthened their confidence and the resiliency to overcome any obstacle. However, the
knowledge and credibility of the other person makes a difference in the level of influence
of this source of self-efficacy (Bandura, 1997). Since the items used to analyze this
particular source deal mainly with the encouragement of teachers, parents, and peers,
19
researchers have yet to analyze the persuasiveness of the message sent by communities,
public figures, media, or society as a whole (Usher & Pajares, 2008).
Physiological and emotional states. The physiological and emotional states have
been assessed in many mathematical research studies as anxiety (Hodges & Murphy,
2009; Lent, Lopez, & Bieschke, 1991; Lent, Lopez, Brown, & Gore, 1996), because all of
the instruments were using variations of the Fennema-Sherman Math Anxiety Scale as
revised by Betz (1978). Other researchers (Matsui et al., 1990; Zeldin & Pajares, 2000)
asked students to assess this construct by thinking of how mathematics made them feel.
In the thought-listing analysis by Lent, Brown, Gover, and Nijjer (1996), responses were
categorized within this construct based on statements of fear, such as math frightens
me. However, it was only responsible for 9% of the responses mentioned by the
participants. As mentioned earlier, Hodges and Murphy (2009) found physiological states
and vicarious experiences were the only two sources that could statistically significantly
predict the self-efficacy of learning mathematics asynchronously. Usher & Pajares (2009)
contend that [a]lthough ones feelings of anxiety may be the most salient form of
psychological arousal in the classroom, particularly in the domain of mathematics, a
measure that includes other forms such as physical arousal and mood would be more
faithful to Banduras (1997) description of this source (p. 91).
Conclusion
Hall and Ponton (2005) suggested that educators have a tendency to teach the
same way with different courses without recognizing the impact it may have on students
with varying levels of self-efficacy. The level of self-efficacy determines the effort
students are willing to put into tasks they find difficult and the resiliency they have in the
20
face of setbacks resulting from those difficult tasks (Bandura, 1997). Thus, teachers of
developmental courses, and more specifically freshmen courses, may want to consider
the self-efficacy of students within their courses. They should strive to create a learning
environment enabling their students to not only master the mathematics concepts during
this transitional year of college, but also to produce students who will become lifelong
learners (Hall & Ponton, 2005).
The studies in this literature review exemplify the various ways in which selfefficacy and the sources of self-efficacy are measured and how probable it is that research
studies could provide conflicting results. Additionally, the measures for self-efficacy
need to take into consideration the specificity of the self-efficacy construct when related
to specific domains, such as mathematics, and adhere to the suggestions posited by
Bandura (1997) when measuring the effects of self-efficacy with the domain specific
performance (Usher & Pajares, 2008). Similarly, more research needs to be done on the
collegiate level utilizing consistent and calibrated measures appropriately studied in
relation to the sources of mathematical self-efficacy.
The lack of current research on the impact of self-efficacy on collegiate
mathematics achievement opens the door for future investigations within areas of
ethnicity and self-efficacy enhancing interventions (Schunk & Pajares, 2002). As the
qualitative research article (Zeldin & Pajares, 2000) showed a connection between
vicarious experiences and verbal and social persuasions with females who chose careers
within the STEM areas, more research should be done to investigate the sources of selfefficacy among other subgroups defined by race, gender, and possibly rural and urban
populations (Usher & Pajares, 2008). Additionally, self-efficacy research related to non-
21
traditional students on the collegiate level needs to be done, especially since research,
according to Cassazza (as cited in Hall & Ponton, 2005), has shown it to be the fastest
growing segment within higher education. Research in these areas will not only add to
the knowledge base regarding self-efficacy and the sources of mathematical self-efficacy,
but also may provide insight into the types of interventions that would be successful and
the areas where those interventions can be applied.
22
23
(see Appendix A for the complete list of categories and selected options). The selections
were made based on the focus of the research study, which pertained to undergraduates of
small, four-year institutions. The selection process resulted in 37 institutions, but seven
were eliminated right away by requiring not-for-profit, four-year or above institutions.
Since 30 institutions were out of the scope of this research study based on time
and resources, the list was processed identically except for selecting only the Very Small
institutions. Of the nine universities produced by this selection process, all but three were
classified as professions plus arts and sciences with some graduate coexistence. So, those
three were eliminated and the remaining six form the target sample used for this research
study. (See Appendix A.)
In January 2012 an email was sent to the Dean of Academic Affairs (or
comparable position) of each of the six institutions selected through the Carnegie
Foundation classifications process mentioned above. One of the institutions chose not to
participate based on a policy within the Registrars Office. Two of the institutions made
no attempt to communicate with the researcher even after several attempts. After the
researcher provided the appropriate information for each of their IRB processes, three of
the original six institutions agreed to participate in the study. Since each of the three
institutions had some type of religious affiliation, the generalizability of the data to the
population may be limited.
All freshmen students enrolled at each participating university in the spring
semester of 2012 were chosen to take part in the study. A contact person at each of the
sample institutions used their database to produce the list of all freshmen students over
the age of 17 (per IRB regulations), including email addresses, and provided the list to
24
the researcher for this study. A total of 474 freshmen students were identified as the
target population. Each of the 474 freshmen students were sent an email (see Appendix
B) containing a brief description of the research study and its importance to the education
community, a link containing written acknowledgement of their rights and assurance of
privacy regarding their information, and a link to the online survey (see Appendix C) for
this research study. Submitting the online survey constituted their consent to participate
in the study.
Instrumentation. The sources of mathematical self-efficacy have been analyzed
by various means as outlined in the review of literature chapter of this study. However,
four specific scales (Lent, Lopez, & Bieschke, 1991; Matsui, Matsui, & Ohnishi, 1990;
Ozyurek, 2005; Usher & Pajares, 2009) have been developed and used more consistently
within the research. Since this research study focused on analyzing the influence of
student characteristics on the four theorized sources of mathematical self-efficacy, it was
important to select an instrument that closely aligned with theory and had been validated
in other research studies.
Matsui, Matsui, and Ohnishi (1990) developed their scale for the sources of
mathematical self-efficacy using the self-reported grades of the students as their mastery
experience score. However, this does not correspond with the theoretical nature of
Banduras mastery experience construct. Mastery experience refers to the manner in
which an individual cognitively processes previous successes and failures. When students
only report their grades, it does not analyze how the grade affected their competence in
mathematics. Since one student may view a C in a course as good and another could view
25
it as bad, then their grades would not be an accurate indicator of their level of selfefficacy (Usher & Pajares, 2009),
Ozyurek (2005) developed a measurement for the sources of mathematics-related
self-efficacy referred to as Math-inform. The Math-inform consisted of only three sources
of self-efficacy, because the first factor contained items related to both mastery
experience and social persuasion. It was not apparent as to why those two constructs were
combined, as they are theorized by Bandura to represent completely different constructs.
Additionally, the instrument used a 4-point Likert scale, which is not sensitive enough to
account for the nuances within cognitive processing (Bandura, 2006).
Lent, Lopez and Bieschke (1991) developed a scale to analyze the four sources of
mathematical self-efficacy of college students. The 40-item instrument was divided into
ten questions per source. All of the sets of questions were designed by them, except for
the questions pertaining to the physiological and emotional states. They used the
Fennema-Sherman Math Anxiety Scale revised by Betz (1978) to analyze that particular
construct. Even though their instrument is more theoretically aligned than the previous
two, the 5-point Likert scale is still not sensitive enough according to Bandura (2006).
Usher and Pajares (2009) developed the Sources of Middle School Mathematics
Self-Efficacy Scale through a 3-phase process. The first phase began with a 6th grade
focus group to determine the understandability of the wording of the instrument. It was
then used during a research study of 1111 middle school students. After conducting an
analysis with the first scale, modifications were made and, during phase two, it was
presented to 824 middle school students. More analysis resulted in more modifications.
Before the instrument was used again during the third phase, the authors submitted their
26
items to experts within the social cognitive theory field (Bandura, Zimmerman, and
Schunk) for feedback. Based on the feedback from the experts more modifications were
made, which resulted in an instrument containing 73 items at the beginning of the final
phase. However, through revisions based on various types of analysis during the final
phase, the official Sources of Middle School Mathematics Self-Efficacy Scale has 24
items consisting of six items per source. Each of the source sections had Cronbachs
alpha reliability coefficients above 0.80 indicating that over 80% of the variance in the
total score for each source of mathematical self-efficacy is shared within the six specific
items on the scale (Warner, 2008). More specifically, the Cronbachs alpha coefficients
for each source was 0.88 for mastery experience, 0.84 for vicarious experience, 0.88 for
social persuasions, and 0.87 for physiological state (Usher & Pajares, 2009).
Comparing the correlation between the sources measures and self-efficacy
outcomes to those obtained in previous research studies of the sources reveals that the
measures created in this study are not only sound, but demonstrate greater predictive
utility than have past measures (Usher & Pajares, 2009, p. 97). For this reason, as well
as the desire to use a valid and calibrated instrument to help further the research on the
sources of mathematics self-efficacy, the Sources of Middle School Mathematics SelfEfficacy Scale was chosen for this research study. It was adapted to be used with collegelevel freshmen students (see Appendix C) by making two cosmetic changes:
Question #2 Seeing kids do better than me in math helps me do better in math.
(kids was replaced by students.)
Question #10 I got good grades on my last report card.
(on my last report card was replaced by in my last math class.)
27
The Sources of Middle School Mathematics Self-Efficacy Scale used a six-point Likert
scale designed for middle school students where the choices were F Definitely False, F
Mostly False, F A little bit False, T A little bit True, T Mostly True, and T
Definitely True. The researcher choose to use the exact same Likert scale for this
research study, because the choice of false and true seemed more appropriate for the
items in the scale than the standard agree and disagree. The Likert scale was converted
to a number from 1 to 6, with 1 representing Definitely False and 6 representing
Definitely True.
Pilot study. A pilot study of 20 upperclassmen at a small, liberal arts institution in
Virginia was conducted in January 2012 to test the instrument for the research study and
to make any necessary adjustments to the instrument and/or the instructions of the
instrument. Based on the results of the pilot study, it was evident that students were
having trouble inputting their high school and college GPA. Changes in the demographic
section of the survey were made to accommodate this situation. Instead of asking students
to input their GPA specifically, they were asked to select the radio button that aligned
with their GPA (see Appendix C). Even though this changed the nature of the variable
from a continuous to discrete variable, the researcher believed it was more important to
have those two data items available for analysis than risk losing out on the information
altogether.
Procedures. In the spring of 2012, all of the college freshmen students (474) in
the three universities were asked to participate in the research study through an email
describing the study and its importance to the educational community. The email (see
Appendix B) contained two Internet links: 1) a website that provided all of the
28
29
immediately separated from the other information and stored in a separate Excel file. The
remaining data was used throughout this research study with no link to any of the
participants. Only 79 students provided their email addresses for the random drawing.
Using a random number generator, the researcher obtained 62 and 34 as the two winners
of the $50 e-gift cards. The researcher sent an email to the participants associated with
those email addresses to thank them again for their participation and to inform them of
their winnings. Since the researcher allowed the students to submit their responses
anonymously, it was not possible to identify the participants based on their institution.
Data Analysis
Measures. The following list contains the variables used for this study.
Dependent Variables
Mastery Experience (ME)
Vicarious Experience (VE)
Verbal and Social Persuasions (VSP)
Physiological and Emotional States (PES)
Vicarious Experience Adult models (VEadult)
Vicarious Experience Peer models (VEpeer)
Vicarious Experience Self model (VEself)
Independent Variables
Gender (Gender) Two levels (Male and Female)
Race (Race) Three levels (African American, Caucasian, and Other)
Sexual Orientation (SexOr) Three levels (Bisexual, Heterosexual, and
Homosexual)
30
31
2008, p. 315). In other words, a factorial ANOVA would determine whether the
combination of gender and race (African American females, Caucasian males, etc.)
interact to create statistically significant mean differences on the mastery experience
dependent variable. This analysis would be performed in SPSS by selecting ANALYSIS
GENERAL LINEAR MODELS UNIVARIATE.
The purpose of a multivariate analysis of variance (MANOVA) would be to
determine whether groups formed by two or more independent variables had statistically
significant mean differences on multiple dependent variables (Warner, 2008). Warner
(2008) goes on to state that the null hypothesis of a MANOVA would correspond with
the assumption that when the scores on all p of the Y outcome variables are considered
jointly as a set, taking intercorrelations among the Y variables into account, the means for
this set of p outcome variables do not differ across any of the populations that correspond
to groups in the study (Warner, 2008, p. 702). In other words, a MANOVA would
determine whether the combination of the independent variables interact to create
statistically significant mean differences on the dependent variables taking
intercorrelations into account. This analysis would be performed in SPSS by selecting
ANALYSIS GENERAL LINEAR MODELS MULTIVARIATE.
The researcher chose to begin the analysis by performing a MANOVA using all
of the independent and dependent variables to determine if there were any statistically
significant differences in the four sources of mathematical self-efficacy taking their
intercorrelations into account. Once the MANOVA was not found to have any
statistically significant results and since, the four theorized sources of mathematical selfefficacy address distinct sources of information (Lent, Lopez, Brown, & Gore, 1996),
32
which can be analyzed separately, the researcher chose to perform factorial ANOVAs for
each of the dependent variables. Furthermore, a Confirmatory Factor Analysis (CFA) was
conducted to test whether the questions contained within the survey were still
corresponding to the source they were intending to analyze. The CFA was conducted
using the AMOS 20 program. All other analyses were conducted using SPSS 20.
Validity
Threats to the reliability and validity of the survey instrument are possible, but
measures were taken to reduce that possibility. Only two minor changes were made to the
Usher & Pajares (2009) survey (see Appendix C) to make it more appropriate for
college-age students. The changes were cosmetic so it was doubtful that it would have
affected the validity of the survey instrument. Threats to reliability, however, are higher
because of the low response rates of the participants.
Out of the 474 emails sent to the freshmen students at the three participating
institutions, 106, or 22.3%, responded. The small sample size and low response rate can
result in a response bias where the responses do not accurately reflect the views of the
sample and the population (Cresswell, 2008, p. 403). By comparing the combined
statistical data (National Center for Education Statistics, 2011) of the three institutions
participating in the research study (11.9% African American, 65.9% Caucasian, 22.2%
other races, 57.8% females and 42.2% males) with the combined statistical data of the
sample (22.6% African American, 61% Caucasian, 16% other races, 50% females and
50% males), the data appears to be comparable with the exception of a higher percentage
of males and African American participants than would be reasonably expected.
33
Since the researcher works at one of the participating institutions with a larger
percentage of students within those two demographic areas, it is probable that her name
was known to the participants and generated more participation from that institution and
specifically those demographic groups. However, all participating students were
informed that their responses would not be known to the researcher, which lessens the
possibility of response bias. Additionally, the responses of the participants were analyzed
as one whole group without breaking down the analysis per institution, which would also
lessen the issues with reliability. However, since the researcher does not know what
percentage of the students attended each of the three institutions, then the overall results
may be unintentionally skewed. Even though the researcher took every opportunity to
make sure the data collected did not have issues with reliability and validity, they are not
completely negated. This combined with the small sample size will limit the
generalizability of the results.
34
35
Therefore, the missing data did not significantly impact the overall score of the source of
mathematical self-efficacy per student. Additionally, during the process of performing the
statistical analysis using SPSS, pairwise deletion was chosen over listwise deletion due to
the small sample size and the need to include as much data as possible within each
analysis.
Demographics
The 102 freshmen in the study are categorized by the following demographics:
49% female, 51% male, 22.5% African American, 62.7% Caucasian, 14.7% other race,
88.2 % heterosexual, 2.9% bisexual, 3.9% homosexual, 39.2% rural hometowns, 27.5%
urban hometowns, and 33.3% suburban hometowns. Unfortunately, the data collected
from the participants regarding sexual orientation (see Figure 4.1) did not provide enough
difference within the sample to allow for this particular characteristic to be used within
any of the analyses.
3.9
4.9
2.9
Bisexual
Heterosexual
Homosexual
No Response
88.2
36
3.00 3.49, 5: 3.50 3.99, 6: 4.00 or above) from which to select their appropriate GPA.
The high school GPA (HSGPA) did not have any student in Level 1, only 5 in Level 2,
and only 8 in Level 6 (see Figure 4.2). The college GPA (COLGPA) had only 6 in Level
1 and 3 in Level 6 (see Figure 4.3).
4.9
7.8
Below 2.00
2.00 - 2.49
20.6
2.50 - 2.99
32.4
3.00 - 3.49
3.50 - 3.99
34.3
4.00 or above
2.9
5.9
25.5
Below 2.00
11.8
2.00 - 2.49
2.50 - 2.99
3.00 - 3.49
24.5
3.50 - 3.99
26.5
4.00 or above
37
Level 2 (3.00 3.49), and the old Levels 5 and 6 became the new Level 3 (3.50 or above)
(see Figure 4.4). For the college GPA the old Levels 1 and 2 became the new college
GPA (COLGPA2) Level 1 (Below 2.50), the old Level 3 became the new Level 2 (2.50
2.99), the old Level 4 became the new Level 3 (3.00 3.49), and the old Levels 5 and 6
became the new Level 4 (3.50 or above) (see Figure 4.5).
25.5
Below 3.00
40.2
3.00 - 3.49
3.50 or above
34.3
28.4
17.6
Below 2.50
2.50 - 2.99
3.00 - 3.49
3.50 or above
24.5
26.5
38
were small compared to the other letter grades. In order to make the levels of the variable
more equal, to address any outliers or abnormalities, and to associate a higher letter grade
with a higher number value (comparable to a standard GPA score), the variable was
reclassified as classgrd with the following levels: Level 4 A grade, Level 3 B grade,
Level 2 C grade, Level 1 D or F grade. The students who withdrew from the course
were eliminated from the new variable, which leaves classgrd with a total of 70 students
(see Figure 4.6).
21.4
28.6
18.6
31.4
A
B
C
D
Figure 4.6 Letter grade received in math course taken the previous semester (classgrd)
The sources of mathematical self-efficacy were calculated at this time by
averaging the values of the participants responses on the six questions pertaining to each
source (see Appendix C). Seven of the questions were reversed coded (marked with an
asterisk) to correspond a higher value with a higher sense of self-efficacy based on that
particular source. The reverse coding matched the coding of the questions from the study
by Usher and Pajares (2009). Mastery experience (ME) was calculated using questions 1,
4, 9*, 10, 13, and 19. Vicarious Experience (VE) was calculated using questions 2, 14,
17, 20, 22, and 24. Verbal and Social Persuasions (VSP) were calculated using questions
6, 8, 11, 16, 18, and 23. Physiological and Emotional States (PES) were calculated using
questions 3*, 5*, 7*, 12*, 15*, and 21*. The more stressed, anxious or depressed a
39
student felt regarding mathematics, the lower the score. Students who felt good about the
mathematics course would have a higher physiological and emotional states score than
those who felt bad.
In order to assess whether combining the scores of the particular questions for all
participants in the research study provide a stable and internally consistent measure
(Warner, 2008), the Cronbachs alpha reliability coefficient was calculated for each
source of mathematical self-efficacy: 0.89 for mastery experience, 0.75 for vicarious
experience, 0.94 for verbal and social persuasions, and 0.90 for physiological and
emotional state. The Cronbachs alpha coefficient was higher in this study for mastery
experience, verbal and social persuasions and physiological and emotional states than
Usher and Pajares (2009) original study (0.88, 0.88, and 0.87, respectively), but lower
than their vicarious experience Cronbachs alpha coefficient (0.84). Usher and Pajares
(2009) have acknowledged that a low Cronbachs alpha coefficient has been an issue
with this particular construct in the past; however, it was still higher in this research study
than other studies (Hodges & Murphy, 2009; Lent, Lopez, & Bieschke, 1991; Matsui,
Matsui, & Ohnishi, 1990).
Usher and Pajares (2009) contend that the low Cronbachs alpha reliability
coefficient of the vicarious experience construct may be a result of research studies
focusing on either peer role models or adult role models, but not both. Within their study
the vicarious experience construct was separated into three subcategories based on the
type of role model: adult, peer and self. Vicarious experience adult refers to the average
of questions 14 and 24. Vicarious experience peer refers to the average of questions 2
and 20. Vicarious experience self refers to the average of questions 17 and 22. Since the
40
Cronbachs alpha reliability coefficient was lower in this study than Usher and Pajares
(2009), the Cronbachs alpha reliability coefficient was analyzed on the three
subcategories as well. The Cronbachs alpha coefficient of reliability was lower for each
of the subsections of the vicarious experience construct (VEadult: = 0.52, VEpeer: =
0.51, and VEself: = 0.62), which implied that all independent items within each
subsection are necessary for the overall reliability of vicarious experience.
After verifying the internal consistency reliability for each of the sources (ME:
= 0.89, VE: = 0.75, VSP: = 0.94, and PES: = 0.90) based on the four groups of six
questions, it was important to determine whether the responses to those questions were
still corresponding to their specified latent variable as established by Usher and Pajares
(2009) through confirmatory factor analysis (CFA).
Confirmatory Factor Analysis
Thomas Schmitt (2011) provided an overview of the various methods available to
researchers interested in performing factor analysis: exploratory factor analysis (EFA)
and confirmatory factor analysis (CFA). According to Schmitt, maximum likelihood
(ML) is one of the most commonly used statistical estimation methods. ML requires at
least 5 participants per item, which would mean at least 120 participants for the 24-item
survey used in this research study, to establish a higher level of reliability per item.
Additionally, ML requires continuous, normally distributed data, which is not found on
survey questionnaires with a Likert scale, such as the one used in this study. Even though
the researcher conducted the ML analysis to verify the issues related to using the analysis
on ordinal data, such as excessive multivariate kurtosis, the results are not included
within this study. Figure 4.7 presents only the model designed for the ML analysis.
41
Figure 4.7 Confirmatory Factor Analysis Model for Maximum Likelihood and Bayesian
estimation of the twenty four survey questions
Additionally, Schmitt (2011) provides robust least squares (LS), robust weighted
least squares (WLS) and Bayesian estimation as the alternative CFA approaches for
ordinal data with two to five categories. Muthen & Asparouhov (2010) proposes applying
the Bayesian approach to factor analysis and structural equation modeling (SEM) because
current analyses using maximum likelihood (ML) and likelihood-ratio 2 testing apply
unnecessarily strict models to represent hypotheses derived from substantive theory (p.
3). In other words, theoretically-based models are often rejected based on the restrictive
limitations of the ML approach (i.e. large sample size of at least 5 participants per item,
normal distributions). Thus, Bayesian estimation was chosen as the alternative method
42
for this research study because it does not require normally distributed, continuous data
from large sample sizes.
Bayesian estimation has not been used widely in social-psychological research
even though it has been around since the 18th century (Arbuckle, 2007; Byrne, 2009).
One reason for its limited appeal to researchers outside of the statistical field may be that
Bayesian analysis appears to be difficult to perform (Muthen & Asparouhov, 2010) and
the computational software to perform the analysis have been unavailable (Arbuckle,
2007).
In the Bayesian approach any quantity that is unknown, such as the four sources
of self-efficacy, is viewed as a random variable and assigned a joint probability
distribution (Arbuckle, 2007). This distribution combines what is known about a
particular parameter, called the prior distribution, with the data-based evidence by the
well-known Bayes Theorem, which results in an updated distribution known as the
posterior distribution (Arbuckle, 2007). In other words, the prior distribution would be
based on prior studies or substantive theory, such as the model developed by Usher &
Pajares (2009) and used for this research study (see the model in Figure 4.7). The databased evidence is the responses of the students to the 24-item questionnaire. The
Bayesian analysis will determine a posterior distribution and a posterior predictive pvalue (PPP) of model fit, where the range of the p-value is between 0 and 1 with a pvalue around 0.5 indicating an excellent fit (Muthen & Asparouhov, 2010). Model
testing is carried out using posterior predictive checking which is found to be less
sensitive than likelihood-ratio 2 testing to ignorable degrees of model misspecification
(Muthen & Asparouhov, 2010, p. 3).
43
Using AMOS 20, the Bayesian approach was applied to this research study
yielding a posterior predictive p-value of 0.50, DIC = 3159.84, and effective number of
parameters = 73.37, which means it was an excellent fit. Once the model fit was
established, the research study continued with an analysis of the results of the survey
items.
Data Analysis
Descriptive statistics. Before data analysis of the survey items began, the
descriptive statistics of the four sources of mathematical self-efficacy and the three
subcategories of vicarious experience were calculated (see Table 4.1). Since the absolute
value of the skewness and kurtosis of each dependent variable was less than 1, the data
was considered normally distributed and parametric tests were used for the analysis.
Table 4.1
Descriptive statistics for the seven dependent variables
Mean
SD
Variance
Skewness
Kurtosis
Mastery
Experience
4.110
1.185
1.405
-.516
-.398
Vicarious
Experience
3.804
.966
.934
-.014
-.554
Verbal &
Social Persuasions
3.564
1.404
1.972
-.093
-.899
Physiological
& Emotional States
4.050
1.250
1.564
-.366
-.534
Vicarious
Experience Adult
3.971
1.128
1.271
-.427
.055
Vicarious
Experience Peer
3.765
1.183
1.400
-.321
-.263
Vicarious
Experience Self
3.691
1.305
1.703
-.061
-.783
44
Effect sizes, power, and Type I error value. Several different effect size indexes
(Cohens d, eta squared (2), partial-eta squared, and R2) are used within research studies
to estimate the proportion of the variance in the scores of the dependent variable based on
the independent variables (Warner, 2008). When the researcher performed the factorial
ANOVAs in SPSS, the output generated the R2 and partial-eta squared effect sizes. The
R2 effect size provides the proportion of variance based on the combination of all of the
independent variables and can be used as an index of the strength of a linear relationship
(Warner, 2008), which would be more appropriate for regression analysis or multivariate
analysis of variance. The partial-eta squared effect size provides the proportion of the
remaining variance in the scores of the sources of mathematical self-efficacy when the
other independent variables and their interactions have been removed (Warner, 2008).
The researcher chose to include the partial-eta squared effect size in the factorial
ANOVA tables in this research study, because it would provide the amount of variance
within the means associated with each individual independent variable when all of the
other variables were held constant.
Additionally, the researcher chose to include the 2 effect size also within the
analysis of this research study, because it can be used to analyze the statistical power of
the results. Statistical power is the probability of obtaining a test statistic large enough to
reject H0 when H0 is false (Warner, 2008). Generally, researchers would like to obtain
80% statistical power. Cohen (as cited in Warner, 2008) established a table to determine
the statistical power from the effect size. Cohen established three levels of effect based
on the d-value (d=.20, small; d=.50, medium; and d=.80, large) and also provided the
corresponding 2 values (.010, .059, and .138, respectively) (Warner, 2008). Warner
45
(2008) also presents Cohens table that provides the sample size necessary to obtain the
desired statistical power based on the value of 2. Since the 2 effect size was not part of
the factorial ANOVA results, the researcher calculated the effect size for each variable by
dividing the sum of squared deviations of the scores of the individual variable based on
the mean of that variable by the sum of squared deviations of all of the scores based on
the overall model (i.e. SSA/SSTotal) (Warner, 2008).
Additionally, the researcher had to determine what percentage of error would be
allowed within the study. Since Cohens statistical power table (as cited by Warner,
2008) was based on the researcher using the standard = .05, two-tailed criterion for
significance, the researcher chose to select = .05 as the desired Type I error risk for the
entire study. In other words, there is a 5% chance that the mean differences found, if any,
between the groups of students would not occur naturally within the population of the
students. Furthermore, when multiple hypothesis testing procedures are used on the same
dependent variable within a study, an adjustment must be made to account for the
increased risk of a Type I error occurring somewhere within the analysis (Huck, 2000).
Since the data analysis consisted of performing two separate factorial ANOVAs on each
dependent variable based on the student characteristics and the performance variables,
then the Bonferroni correction rate of .05/2 (.025) was utilized per the Bonferroni
technique (Huck, 2000).
Mean differences. As mentioned previously, the researcher chose to perform a
MANOVA on all of the independent and dependent variables to determine whether there
were statistically significant differences in the sources of mathematical self-efficacy
when their intercorrelations were taken into account. Based on the MANOVA, none of
46
the independent variables nor their interactions were found to be statistically significant.
Additionally, the changes between R2 and the adjusted R2 for each of the sources of selfefficacy indicated a huge loss in power (e.g. vicarious experience: R2 = .915, adjusted R2
= -.236). Since the four theorized sources of mathematical self-efficacy address distinct
sources of information (Lent, Lopez, Brown, & Gore, 1996), which can be analyzed
separately, and due to the loss in power of the MANOVA, the researcher felt justified in
performing factorial ANOVAs on each of the sources of self-efficacy.
Factorial ANOVAs were important in this research study because they provided
information regarding interaction effects of independent variables on the dependent
variables. However, with a sample size of only 102 students and six independent
variables with two or more levels each, factorial ANOVAs posed a problem. In order to
analyze data, the number of students within each subcategory of the specific factorial
ANOVA needed to be somewhat similar. When one subcategory did not contain any
students (i.e. no males of other races) or only a few compared to other subcategories,
issues with analysis arose. In order to proceed with the factorial ANOVAs for each
dependent variable, Race was re-classified with only two levels: African-American 1
and Caucasian 2.
In addition to the assumption of normally distributed scores on the dependent
variables, ANOVAs require equal variances across the populations. The Levenes test of
homogeneity assesses the null hypothesis that the variances are equal, which means that
the researcher hopes to find a non-significant test statistic result (Warner, 2008). The
Levenes test was non-significant for all but one factorial ANOVA (vicarious experience
self with student characteristics as the independent variable). The researcher chose to
47
run three separate one-way ANOVAs for each of the student characteristic variables
(gender, race and hometown location) on the vicarious experience self dependent
variable. The three one-way ANOVAs all passed the Levenes test of homogeneity.
Furthermore, for each of the factorial and one-way ANOVAs, the researcher chose to
include a Tukey honestly significant difference (HSD) post-hoc test, which shows all
possible pairwise comparisons of group means and reports whether each of these
comparisons is significant (Warner, 2008, p. 241). In other words, the Tukey HSD test
will identify which of the levels of the variable, if any, have statistically significant mean
differences.
Mastery experience. Each of the two factorial ANOVAs analyzed on the
dependent variable, mastery experience, had one statistically significant result at the
adjusted p = .025 level, but no interaction effects. Hometown location (see Table 4.2) and
the letter grade received in the mathematics course from the previous semester (see Table
4.3) were found to be statistically significant independent variables.
Table 4.2
2x2x3 Factorial ANOVA on Mastery Experience with student characteristics as
independent variables
Source
Corrected Model
Gender
Home
Race
Gender * Home
Gender * Race
Home * Race
Gender * Home * Race
Error
Total
Corrected Total
*p<.025
Type III SS
23.69
.02
13.02
.32
4.02
1.33
.84
4.76
94.00
1620.63
117.70
df
MS
11
1
2
1
2
1
2
2
75
87
86
2.15
.02
6.51
.32
2.01
1.33
.42
2.38
1.25
48
F
1.72
.02
5.19
.26
1.60
1.06
.34
1.90
p
.085
.894
.008*
.615
.208
.306
.715
.157
Partial 2
.20
.00
.12
.00
.04
.01
.01
.05
2
.20
.00
.11
.00
.03
.01
.01
.04
Type III
Partial
df
MS
60.07
28
21.45
2.67
.002
.66
.66
HSGPA2
2.22
1.11
1.38
.265
.07
.02
COLGPA2
2.27
.76
.94
.431
.07
.02
20.29
6.76
8.40 <.001*
.39
.22
HSGPA2 * COLGPA2
2.87
.48
.60
.732
.08
.03
HSGPA2 * classgrd
7.11
1.19
1.47
.213
.19
.08
COLGPA2 * classgrd
2.48
.41
.51
.795
.07
.03
.55
.28
.34
.713
.02
.01
Error
31.39
39
.81
Total
1247.98
68
91.46
67
Corrected Model
classgrd
Corrected Total
SS
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade in
mathematics course
*p < .025
49
The letter grade students received in their mathematics course from the previous
semester was determined to be statistically significantly influential on the mastery
experience construct: F (3, 39) = 8.40, p < .025. This corresponds to a large effect size of
2 = .22, which means that about 22% of the variance in the mastery experience scores
was predictable from the letter grade the students received in their mathematics course
from the previous semester. The partial-eta squared value indicates that the letter grade
from the mathematics course the previous semester accounts for 39% of the variance in
the mastery experience scores when all of the other variables are held constant. The
Tukey HSD post-hoc test showed that students who self-reported an A in their
mathematics course in the previous semester (M = 5.08, SD = 0.56) had a statistically
significantly higher mastery experience score than those students who had a C (M = 3.52,
SD = 1.17) or those who had a D or below (M = 3.15, SD = 1.11). Additionally, the
Tukey HSD post-hoc test showed that students who self-reported a B in their
mathematics course from the previous semester (M =4.30, SD = 0.84) had a statistically
significantly higher mastery experience score than those who had a D or below (M =
3.15, SD = 1.11).
Warner (2008) explains how a significant F-test could yield non-significant
Tukey HSD results for all level comparisons, because the Tukey HSD test requires a
slightly larger difference between means to achieve significance (p. 247). However, the
researcher could not find a reasonable explanation as to why a non-significant F-test,
such as the one for the students college GPA, did have statistically significant
differences in the levels of the variable on the Tukey HSD test at the adjusted p-value.
According to the Tukey HSD test, students who reported a college GPA of 3.5 or above
50
Type III SS
df
MS
Partial
2
Corrected Model
17.63
11
1.60
1.97
.044
.22
.22
Gender
.54
.54
.66
.419
.01
.01
Startseite
5.22
2.61
3.20
.046
.08
.07
Race
.41
.41
.51
.479
.01
.01
Gender * Home
2.46
1.23
1.51
.228
.04
.03
Gender * Race
.32
.32
.40
.530
.01
.00
Home * Race
1.13
.56
.69
.505
.02
.01
2.61
1.30
1.60
.209
.04
.03
Error
61.16
75
.82
Total
1334.31
87
78.79
86
Corrected Total
Additionally, even though the F-test was not significant, the Tukey HSD test found that
students who self-reported an A (M = 4.28, SD = 0.84) in their mathematics course taken
in the previous semester had statistically significantly higher vicarious experience scores
51
than those students who had a D or below (M = 3.25, SD = 0.80). The researcher chose to
include this statistical difference, but encourages others to interpret it with caution.
Table 4.5
3x4x4 Factorial ANOVA on the dependent variable, vicarious experience, with
performance variables as the independent variables
Source
Type III SS
df
MS
Partial
2
Corrected Model
24.15
28
.86
.87
.644
.39
.39
HSGPA2
1.09
.55
.55
.580
.03
.02
COLGPA2
1.82
.61
.61
.610
.05
.03
classgrd
8.80
2.93
2.97
.044
.19
.14
HSGPA2 * COLGPA2
2.01
.34
.34
.912
.05
.03
HSGPA2 * classgrd
3.51
.59
.59
.735
.08
.06
.97
.16
.16
.985
.03
.02
1.54
.77
.78
.465
.04
.02
Error
38.59
39
.99
Total
1014.50
68
62.74
67
COLGPA2 * classgrd
HSGPA2 * COLGPA2 * classgrd
Corrected Total
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics
course
Verbal and social persuasions. The factorial ANOVA on the dependent variable,
verbal and social persuasions, with the student characteristics as the independent
variables did not have any statistically significant results at the adjusted p = .025 level
(see Table 4.6). However, the p-value (.030) of the students hometown environment
warrants further research. Additionally, the factorial ANOVA on the dependent variable,
verbal and social persuasions, was statistically significantly influenced at the adjusted p =
.025 level based on the letter grade received in the mathematics course from the previous
semester (see Table 4.7).
The letter grade students received in their mathematics course from the previous
semester was determined to be statistically significantly influential on the verbal and
52
social persuasions construct: F (3, 39) = 4.37, p < .025. This corresponds to a large effect
size of 2 = .15, which means that about 15% of the variance in the verbal and social
persuasions scores was predictable from the letter grade the students received in their
mathematics course from the previous semester. The partial-eta squared value indicates
that the letter grade from the mathematics course the previous semester accounts for 25%
of the variance in the verbal and social persuasions scores when all of the other variables
are held constant. The Tukey HSD post-hoc test showed that students who self-reported
an A (M = 4.48, SD = .93) in their mathematics course in the previous semester had a
statistically significantly higher verbal and social persuasions score than those students
who had a C (M = 2.53, SD = 1.31) or those who had a D or below (M = 3.06, SD =
1.27). Since students who receive an A in their courses are more likely to earn verbal
praises from family and friends, it seems appropriate for their verbal and social
persuasions scores to be higher than those students who received a C or below.
Table 4.6
2x2x3 Factorial ANOVAs on the dependent variable, verbal and social persuasions, with
the student characteristics as the independent variables
Source
Type III SS
df
MS
Partial
2
Corrected Model
33.74
11
3.07
1.72
.085
.20
.20
.50
.50
.28
.598
.00
.00
13.08
6.54
3.67
.030
.09
.08
.68
.68
.38
.538
.01
.00
Gender * Home
7.31
3.66
2.05
.135
.05
.04
Gender * Race
.98
.98
.55
.461
.01
.01
Home * Race
.91
.46
.26
.775
.01
.01
6.31
3.16
1.77
.177
.05
.04
Error
133.58
75
Total
1242.19
87
167.32
86
Gender
Home
Race
Corrected Total
53
Table 4.7
3x4x4 Factorial ANOVA on the dependent variable, verbal and social persuasions, with
the performance variables as the independent variables
Source
Type III SS
df
MS
Partial
2
Corrected Model
76.78
28
2.74
1.84
.039
.57
.57
HSGPA2
8.23
4.12
2.76
.076
.12
.06
COLGPA2
1.65
.55
.37
.777
.03
.01
19.57
6.52
4.37
.010*
.25
.15
HSGPA2 * COLGPA2
4.62
.77
.52
.793
.07
.03
HSGPA2 * classgrd
8.67
1.45
.97
.459
.13
.06
COLGPA2 * classgrd
5.09
.85
.57
.753
.08
.04
2.68
1.34
.90
.416
.04
.02
Error
58.16
39
1.49
Total
968.17
68
Corrected Total
134.93
67
Classgrd
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics
course
*p < .025
54
Table 4.8
2x2x3 Factorial ANOVA on the dependent variable, physiological and emotional states,
with the student characteristics as the independent variable
Source
Type III SS
df
MS
Partial
2
Corrected Model
15.69
11
1.43
.93
.515
.12
.12
Gender
.04
.04
.03
.871
.00
.00
Startseite
8.46
4.23
2.76
.069
.07
.06
Race
3.40
3.40
2.22
.140
.03
.03
Gender * Home
3.26
1.63
1.07
.349
.03
.02
Gender * Race
.53
.53
.35
.557
.01
.00
Home * Race
.68
.34
.22
.802
.01
.01
.37
.19
.12
.886
.00
.00
Error
114.73
75
1.53
Total
1504.87
87
130.42
86
Corrected Total
The letter grade students received in their mathematics course from the previous
semester was determined to be statistically significantly influential on the physiological
and emotional states construct: F (3, 39) = 5.95, p < .025. This corresponds to a large
effect size of 2 = .20, which means that about 20% of the variance in the physiological
and emotional states scores was predictable from the letter grade the students received in
their mathematics course from the previous semester. The partial-eta squared value
indicates that the letter grade from the mathematics course the previous semester
accounts for 31% of the variance in the physiological and emotional states scores when
all of the other variables are held constant. The Tukey HSD post-hoc test showed that
students who self-reported an A (M = 4.81, SD = 0.96) in their mathematics course in the
previous semester had a statistically significantly higher physiological and emotional
states score than those students who had a C (M = 3.14, SD = 1.21) or those who had a D
or below (M = 3.40, SD = 1.17). The Tukey HSD test also showed that students who self-
55
Type III SS
df
MS
Partial
2
Corrected Model
65.57
28
2.34
1.86
.037
.57
.57
HSGPA2
1.74
.87
.69
.507
.03
.02
COLGPA2
6.09
2.03
1.61
.203
.11
.05
22.55
7.52
5.95
.002*
.31
.20
6.57
1.10
.87
.528
.12
.06
10.52
1.75
1.39
.244
.18
.09
6.79
1.13
.90
.507
.12
.06
.56
.28
.22
.801
.01
.00
Error
49.25
39
1.26
Total
1194.04
68
114.82
67
classgrd
HSGPA2 * COLGPA2
HSGPA2 * classgrd
COLGPA2 * classgrd
HSGPA2 * COLGPA2 * classgrd
Corrected Total
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics
course
*p < .025
Vicarious experience adult role model. Each of the two factorial ANOVAs
analyzed on the dependent variable, vicarious experience adult role models, with
student characteristics (see Table 4.10) and performance variables (see Table 4.11) as the
56
Type III SS
df
MS
Partial 2
11.40
11
1.04
.82
.619
.11
.11
Gender
.03
.03
.02
.877
.00
.00
Startseite
4.32
2.16
1.71
.188
.04
.04
Race
.01
.01
.01
.921
.00
.00
Gender * Home
1.94
.97
.77
.467
.02
.02
Gender * Race
.01
.01
.01
.917
.00
.00
Home * Race
1.15
.58
.46
.636
.01
.01
2.24
1.12
.89
.416
.02
.02
Error
94.62
75
1.26
Total
1462.25
87
106.02
86
Corrected Total
Table 4.11
3x4x4 Factorial ANOVA with the dependent variable, vicarious experience adult role
models, with the performance variables as the independent variables
Source
Corrected Model
Type III SS
df
MS
Partial 2
36.74
28
1.31
1.09
.394
.44
.44
.85
.42
.35
.706
.02
.01
2.66
.89
.74
.536
.05
.03
12.17
4.06
3.37
.028
.21
.15
7.19
1.20
1.00
.441
.13
.09
14.42
2.40
2.00
.089
.24
.17
3.24
.54
.45
.842
.07
.04
.71
.36
.30
.746
.02
.01
Error
46.89
39
1.20
Total
1120.25
68
83.63
67
HSGPA2
COLGPA2
classgrd
HSGPA2 * COLGPA2
HSGPA2 * classgrd
COLGPA2 * classgrd
HSGPA2 * COLGPA2 * classgrd
Corrected Total
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics course
57
Vicarious experience peer. Each of the two factorial ANOVAs analyzed on the
dependent variable, vicarious experience peer role models, with student characteristics
(see Table 4.12) and performance variables (see Table 4.13) as the independent variables
had no statistically significant results at the adjusted p-value of .025. However, the pvalue (.035) for the interaction effects of gender and hometown environment with a
medium effect size of 2 = .07 warrants further research.
Table 4.12
2x2x3 Factorial ANOVA with dependent variable, vicarious experience peer role
model, with student characteristics as the independent variables
Source
Corrected Model
Type III SS
df
MS
Partial 2
25.53
11
2.32
1.93
.049
.22
.22
Gender
.01
.01
.01
.933
.00
.00
Startseite
6.58
3.29
2.73
.072
.07
.05
Race
.05
.05
.04
.838
.00
.00
Gender * Home
8.45
4.23
3.51
.035
.09
.07
Gender * Race
.01
.01
.01
.915
.00
.00
Home * Race
1.46
.73
.61
.547
.02
.01
5.01
2.51
2.08
.132
.05
.04
Error
90.32
75
1.20
Total
136.00
87
Corrected Total
115.85
86
Table 4.13
3x4x4 Factorial ANOVA with the dependent variable, vicarious experience peer role
model, with performance variables as the independent variables
Source
Corrected Model
Type III SS
df
MS
Partial 2
38.15
38
1.36
.92
.587
.40
.40
.89
.44
.30
.743
.02
.01
COLGPA2
1.00
.33
.23
.878
.02
.01
classgrd
8.07
2.69
1.82
.160
.12
.08
HSGPA2 * COLGPA2
9.46
1.58
1.06
.400
.14
.10
HSGPA2 * classgrd
6.60
1.10
.74
.619
.10
.07
HSGPA2
58
3.32
.55
.37
.892
.05
.03
3.45
1.73
1.17
.323
.06
.04
Error
57.80
39
1.48
Total
1018.75
68
95.95
67
Corrected Total
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics course
Vicarious experience self role model. As was mentioned earlier in this chapter,
the subcategory of the vicarious experience construct pertaining to the self generated role
model failed the homogeneity assumption for the factorial ANOVA using the student
characteristics as the independent variables. The researcher chose to run three one-way
ANOVAs on the vicarious experience self dependent variable for each of the student
characteristics, gender, race, and hometown location, (see Tables 4.14, 4.15, and 4.16,
respectively) Since each of the factorial ANOVAs on the student characteristics was
using the adjusted p-value of .025 and the homogeneity assumption failed for this
subcategory of the vicarious experience construct, the researcher divided the error rate of
.025 by three for each of the one-way ANOVAs to keep the overall Type I error rate at
0.5. Thus the .025 p-value was further adjusted to a p-value of .025/3, or approximately
.008, based on the Bonferroni correction technique.
Gender was the only independent variable (including the performance variables in
the factorial ANOVA see Table 4.17) found to statistically significantly influence the
vicarious experience - self score: F(1, 100) = 9.03, p < .008. This corresponds with a
medium effect size of 2 = .08, which means that 8% of the variance in the vicarious
experience - self scores was predictable from the gender of the students. Male students
(M = 4.06, SD = 1.20) had a higher vicarious experience - self score than female students
59
(M = 3.31, SD = 1.31). Additionally, the p-value (.016) for race with a medium effect
size of 2 = .07 warrants further research on how race influences the subcategory of the
vicarious experience construct pertaining to the self generated role model.
Table 4.14
One-Way ANOVA on the dependent variable, vicarious experience self role model, with
the independent variable, gender
Source
SS
Between Groups
14.25
14.25
157.77
100
1.58
172.02
101
Within Groups
Total
* p < .008, 2 = .08
df
MS
F
9.03
p
.003*
Table 4.15
One-Way ANOVA on the dependent variable, vicarious experience self role model, with
the independent variable, race
Source
Between Groups
SS
df
MS
9.68
9.68
Within Groups
136.57
85
1.61
Total
146.25
86
6.03
.016
= .07
2
Table 4.16
One-Way ANOVA on the dependent variable, vicarious experience self role model, with
the independent variable, hometown environment
Source
Between Groups
SS
df
MS
5.66
2.83
Within Groups
166.36
99
1.68
Total
172.02
101
= .03
2
60
1.68
.191
Table 4.17
3x4x4 Factorial ANOVA with the dependent variable, vicarious experience with self role
model, with the performance variables as the independent variables
Source
Type III SS
df
MS
Partial
2
Corrected Model
47.37
28
1.69
.99
.502
.42
.42
HSGPA2
10.94
5.47
3.21
.051
.14
.10
2.24
.75
.44
.727
.03
.02
12.58
4.19
2.46
.077
.16
.11
HSGPA2 * COLGPA2
5.77
.96
.56
.756
.08
.05
HSGPA2 * classgrd
7.16
1.19
.70
.652
.10
.06
COLGPA2 * classgrd
4.27
.71
.42
.863
.06
.04
2.77
1.38
.81
.452
.04
.02
Error
66.55
39
1.71
Total
1014.75
68
113.92
67
COLGPA2
classgrd
Corrected Total
Note. HSGPA2 high school GPA; COLGPA2 college GPA; classgrd letter grade of mathematics
course
Conclusion
This research study determined several substantial results for the student
characteristics and the performance variables on the four sources of mathematical selfefficacy and the subcategories of the vicarious experience construct at the adjusted pvalue of .025 based on the Bonferroni correction technique with medium to large effect
sizes.
61
The grade received in the math course the previous semester was found to have a
statistically significant impact on three of the four sources of mathematical selfefficacy at the adjusted p-value of .025.
Based on all three performance variables (high school GPA, college GPA, and
letter grade of the mathematics course taken the previous semester), only the letter
grade had a statistically significant influence on several of the dependent
variables.
Race and high school GPA did not have a statistically significant influence on any
of the dependent variables at the adjusted p-value of .025, but some of their pvalues warrant further research.
College GPA did not have a statistically significant influence on any of the
dependent variables.
Even though this research study did not determine any statistically significant
interaction effects between any of the independent variables on any of the
dependent variables, the p-value (.035) of the interaction of gender and hometown
environment on the peer role model subcategory of the vicarious experience
construct warrants further research.
62
63
64
65
Future research should focus on how African-American mathematical role models impact
the sources of mathematical self-efficacy.
Hometown environment. A students hometown environment (rural, urban or
suburban) could have an impact on the four sources of self-efficacy based on the types of
role models present within the community, percentage of parental involvement, and types
of verbal and social persuasion from the family unit and community. This student
characteristic was found to be statistically significant within the scope of this research
study at the adjusted rate of .025 on the mastery experience construct. Since this type of
independent variable has not been included in studies pertaining to the sources of
mathematical self-efficacy, the statistically significant results demonstrate a need for
further research. In addition, based on the mean average scores for the four sources of
mathematical self-efficacy, rural and suburban students within this study favored mastery
experience, first; physiological & emotional states, second; vicarious experience, third;
and verbal & social persuasions, last. Whereas, urban students within this study favored
physiological & emotional states, first; mastery experience, second; closely followed by
vicarious experience, third; and then verbal & social persuasions.
Since students would have spent more time acclimating to their own hometown
environment, it makes sense that it would impact various areas of their lives, including
their self-efficacy. However, as a student integrates more into the college environment
and is surrounded by people from various backgrounds, the impact of the hometown
environment on the sources of self-efficacy may begin to lessen. A longitudinal study
focusing on the impact of the hometown environment on the four sources of
mathematical self-efficacy would be informative.
66
High school GPA. Even though this performance variable was not found to be
statistically significant within the scope of this research study at the adjusted rate of .025,
a p-value of .076 with a medium effect size (2 = .06) on the mastery experience
construct and a p-value of .051 with a medium effect size of (2 = .10) on the self
generated role model subcategory of the vicarious experience construct warrants more
investigation. Since a students high school GPA is a performance outcome based on all
courses taken within the high school years, it would seem likely to correspond with the
mastery experience construct. On the other hand, the overall high school performance
outcome does not correspond with how efficacious a student feels regarding his/her
mathematical abilities. Thus, the Sources of Middle School Mathematics Self-Efficacy
Scale (Usher & Pajares, 2009) revised for this research study provides a more accurate
assessment of the informative nature of the sources of self-efficacy than the scale
developed by Matsui, Matsui, and Ohnishi (1990), because their mastery experience
construct was determined by the self-reported high school mathematics grade. Using high
school GPA as a means of measuring mastery experience of mathematics or academics in
general goes against the tenets of Banduras (1997) social cognitive theory.
College GPA. This performance variable was not found to be statistically
significant within the scope of this research study at the adjusted rate of .025 for any of
the dependent variables, which supports Banduras (1997) tenet, that a numerical value,
such as GPA, does not measure the overall efficaciousness of a student. Even though this
student characteristic was not found to be statistically significant on any of the sources of
mathematical self-efficacy, the Tukey HSD test did find statistically significant mean
differences in the scores of the physiological and emotional states construct. Students
67
who had a college GPA of 2.50 2.99 had a lower score than students with a GPA of 3.5
or higher. It is reasonable that a student with a higher college GPA would not feel
anxious or depressed about mathematics. However, some of the students in the study had
not taken a college-level mathematics course, which could be the reason for the mixed
results. Regardless of the results of the Tukey HSD test, the tenets of Banduras (1997)
social cognitive theory does not support using a performance variable as a measure of any
of the sources of mathematical self-efficacy.
Mathematics course grade. The grade received in the mathematics course taken
the previous semester was the only performance variable found to statistically
significantly influence the sources of mathematical self-efficacy. It was found to have a
statistically significant impact on three of the four sources of mathematical self-efficacy
(mastery experience, verbal and social persuasions, and physiological and emotional
states) at the adjusted rate of .025. This performance outcome is more closely associated
with the sources of mathematical self-efficacy, based on its statistical significance with
three of the four sources, than even college GPA, because these students had already
experienced a college-level mathematics course. Even though this variable highly
corresponds with the sources, researchers are cautioned not to use the grade as a means of
assessing mastery experience or any of the other sources, because it does not account for
how a student cognitively interprets the grade received and it goes against the tenets of
the social cognitive theory. Furthermore, a p-value of .028 on the vicarious experience
adult construct, a p-value of .077 on the vicarious experience self construct warrants
more research within this area as well.
68
Sexual orientation. Due to the small sample size and the large percentage
(88.2%) of heterosexual students, it was not possible to analyze this particular student
characteristic within the scope of this research study. The small number of homosexual
and bisexual students in this sample may have been associated with the religious
affiliation of the institutions participating in the research study. However, the lack or
nonexistence of research in this area regarding the sources of mathematical self-efficacy
indicates that more research should be conducted with attention to the sexual orientation
of students, especially since some research (Rahman & Wilson, 2003; Peters, Manning,
& Reimers, 2007) has found statistically significant differences between gender and
sexual orientation on tests of spatial processing, such as the mental rotation task (MRT).
Future research should be conducted on public institutions without religious affiliation to
generate a more diverse sample.
Discussion
The purpose of this descriptive, survey research study was to determine which
student characteristics (race, gender, sexual orientation and hometown location) influence
the sources of mathematical self-efficacy of college freshmen, if any, utilizing a valid
assessment tool aligned with the tenets of the self-efficacy theory. Based on the concept
that personal agency is socially rooted and operates within sociocultural influences,
individuals are viewed both as products and producers of their own environments and of
their social systems (Pajares, 1996, p. 544), the researcher hypothesized that groups of
students would be influenced by similar sources of self-efficacy (i.e. the vicarious
experience score of the participants will be influenced by their race or the verbal and
social persuasion score will be influenced by a students hometown location). This study,
69
utilizing the valid assessment tool designed by Usher and Pajares (2009) slightly
modified for college students, confirmed the hypothesis for two independent variables.
Mastery experience was found to be influenced by the hometown environment
with suburban students scoring higher than urban students. Since the hometown
environment of students has not been analyzed in previous research studies, this
statistically significant result indicates a need for further research on this particular
student characteristic. Additionally, mastery experience, verbal and social persuasions,
and physiological and emotional states was found to be influenced by the grade received
in the mathematics course taken the previous semester. The statistically significant
influence of this performance variable on several of the sources of mathematical selfefficacy is not surprising based on the fact that grades are often used as the sole measure
of ability within the educational system. However, using a students previous grade in a
particular course to analyze the sources of self-efficacy does not correspond with the
tenets of Banduras (1997) social cognitive theory.
In their qualitative study of successful women in STEM career fields, Zeldin and
Pajares (2000) determined that vicarious experience and verbal and social persuasions
were highly influential with their participants. However, gender did not have a
statistically significant impact on any of the four sources of mathematical self-efficacy
within this research study. This difference could be a result of the small sample size
(n=102) of this research study, small sample size of their qualitative research study
(n=15), or the results of their single-gendered study. Since their research study did not
include an analysis of successful males in STEM career fields, it is not possible to state
definitively that women are more influenced by vicarious experiences or verbal and
70
social persuasions. It is possible that males would have answered those same questions in
a similar manner. Additionally, students interested in pursuing STEM careers would be
expected to have higher mathematical self-efficacy than those who are not pursuing those
careers. Since this current research study did not ascertain the interests of the students to
pursue STEM related careers, it is not possible to determine whether this interest
significantly impacted possible gender differences.
Gender was found in the current research study to have a statistically significant
influence on the self-generated role model subcategory of the vicarious experience
construct. Males scored higher than females on this particular dependent variable. Race
and the interaction of gender with hometown environment produced p-values of interest
for at least one of the subcategories as well. However, the low reliability of the three
subcategory items (adult: = 0.52, peer: = 0.51, and self: = 0.62) limits the
interpretability of the findings.
Stevens et al. (2004) determined that prior mathematics achievement was more
influential for Hispanic students than for Caucasian students, which may cause others to
believe that race would have a statistically significant influence on the sources of
mathematical self-efficacy, or at least the mastery experience construct. However, within
the scope of this small research study, race was not found to be statistically significant for
any of the four sources. One reason for this particular difference is the manner in which
Stevens et al. assessed the prior mathematics achievement. They had students self-report
the grade they normally made in mathematics courses instead of analyzing the mastery
experience construct within the frame of the tenets of the social cognitive theory. Race
71
may have an influence on the sources of mathematical self-efficacy, but the small sample
size of the current study could have contributed to the non-significant results.
Even though Usher and Pajares confirmed the validity and reliability of their
Sources of Middle School Mathematics Self-Efficacy Scale within their study (2009),
investigating the impact of student characteristics, such as gender and ethnicity, on the
sources of mathematical self-efficacy was beyond the scope of their study. Thus, the
results within this research study provides the first analysis of the impact of student
characteristics on the four sources of mathematical self-efficacy utilizing their scale, but
slightly adapted for college level students.
The fact that each person is a combination of environmental, personal and
behavioral factors may explain the limited statistically significant results obtained from
this research study and may also explain some of the obstacles mathematics teachers face
while trying to guide students to mathematical success. Mathematics teachers have an
obligation to help enhance the mathematical self-efficacy of all students, especially
during this highly technologically advanced period of history. Focusing on ways to
enhance each of the four sources of self-efficacy during the course of the mathematics
class is one way to enhance the overall mathematical self-efficacy of all students.All
students have a capacity to learn mathematics, but the capacity is influenced by familial,
biological, environmental, and social factors. Teachers must focus on providing the best
social and environmental arenas conducive to learning by making their classrooms a safe
haven for curiosity, exploration and mistakes. No one builds knowledge without some
struggles and errors. Students may be scared about making mistakes or may be
embarrassed to be wrong in front of their peers, but mistakes should be celebrated for
72
their learning potential. Teachers must establish an environment of mutual respect for all
students by modeling respect for all individuals and providing encouragement throughout
the learning process. Within a respectful, student-centered environment, students will feel
free to search for answers through multiple strategies and discover that learning is a lifelong process. By creating a safe haven for curiosity, exploration, celebration of mistakes,
and respect for all individuals, teachers will ultimately help enhance all four of the
sources of self-efficacy.
Conclusion
In Usher & Pajares Sources of Self-Efficacy in School: Critical Review of the
Literature and Future Directions (2008), the only student characteristics within the
critical review pertained to gender, ethnicity and academic level. This research study was
able to add to the literature by discussing the statistically significant results of a new
student characteristic: hometown environment. Additionally, this study introduced the
idea of another new student characteristic, sexual orientation, to the study of the sources
of mathematical self-efficacy, but due to the nature of the participants no analysis was
able to be performed. These two new student characteristics should be investigated
further within a more diverse population from larger, public institutions, because they
may provide further insight into the connection between students and their sources of
mathematical self-efficacy.
In addition, this study provided further insight into the connection of gender, race,
and the performance outcomes (high school GPA, college GPA, and the grade received in
a math course the semester prior to the research study) with the sources of mathematical
self-efficacy. Furthermore, this research study made slight modifications to the Sources
73
of Middle School Mathematics Self-Efficacy Scale (2008) for use with college-level
students and, through the Bayesian estimation approach, confirmed the factor loadings of
the 24-item scale. Future research in secondary and post-secondary education should
continue to analyze the validity and structure of this scale while assessing the impact of
student characteristics on the sources of mathematical self-efficacy.
74
APPENDIX A
Custom Listings Categories and Selections for the Research Study
Undergraduate Instructional Program selected all arts and science plus professional
baccalaureates with some graduate coexistence
1) A&S + Prof/SGC
2) Bal/SGC
3) Prof + A&S/SGC
Graduate Instructional Program selected only single post baccalaureate institutions
1) S-Post bac/Ed
2) S-Post bac/Bus
3) S-Post bac/Other
Enrollment Profile selected institutions with majority or more undergraduate
populations
1) VHU: Very high undergraduate
2) HU: High undergraduate
3) MU: Majority undergraduate
Undergraduate Profile selected all full time four year institutions
1) FT4/I: Full-time four year, inclusive
2) FT4/S/LTI: Full-time four year, selective, lower transfer-in
3) FT4/S/HTI: Full-time four year, selective, higher transfer-in
4) FT4/MS/LTI: Full-time four year, more selective, lower transfer-in
5) FT4/MS/HTI: Full-time four year, more selective, higher transfer-in
Size and Setting selected all small and very small institutions
1) VS4/NR: Very small four year, primarily nonresidential
2) VS4/R: Very small four year, primarily residential
3) VS4/HR: Very small four year, highly residential
4) S4/NR: Small four-year, primarily nonresidential
5) S4/R: Small four-year, primarily residential
6) S4/HR: Small four-year, highly residential
Basic Classifications selected all options pertaining to baccalaureates
1) Bac/A&S: Baccalaureate Colleges Arts and Science
2) Bac/Diverse: Baccalaureate Colleges Diverse Fields
3) Bac/Assoc: Baccalaureate Colleges Associates Colleges
Final filtering stage involved selecting 4-year or above and private, not-for-profit
institutions. The final six institutions selected for participation in this research study:
Alderson Broaddus College, WV; Averett University, VA; Blue Mountain College, MS;
Kansas Wesleyan University, KS; Saint Gregorys University, OK; Wilson College, PA
75
APPENDIX B
Letter to Students
Dear Student,
Your college is one of three institutions participating in a research study pertaining to the
level of confidence in the mathematical abilities of first year college students. Your
assistance with this study will help further this research and hopefully result in providing
colleges and universities with the information to help them strengthen the mathematical
confidence of future students. You are being invited to participate in this study by
submitting your responses to a very quick online survey. Your survey responses will be
returned electronically verifying your submission but with no identifying information. In
the final research documentation, the name of your college will not be included either. If
you agree to participate in this research study, your name will be placed in a drawing for
two separate $50 online gift cards.
Every person is unique and all responses are important to paint a more accurate picture of
the mathematical confidence level of college freshman. Even though your response to
this survey instrument is needed to provide sufficient data for the analysis, your
participation in this study is completely voluntary. You may withdraw from the study at
any time, as well as refuse to answer any questions. Students under the age of 18 are not
allowed to participate in this research study. For further information regarding the
research study including the rights of the participants and the confidentiality of the data
go to [email protected].
The survey should only take about 10 minutes or less of your time. Please answer the
questions as honestly as possible. Once you submit your responses to the questionnaire,
you will be giving your informed consent to participate in this research study.
When you are ready to begin the survey, please click the following link. <URL>
Thanks again for participating!
Tonja M. Locklear
Doctoral Candidate
STEM Education Department
University of Kentucky
76
APPENDIX C
Research Survey
Race:
Sexual Orientation:
Hometown:
area?
Bisexual ____
Caucasian ____
Other ____
Heterosexual _____
Homosexual ____
Is your home in a more rural (country) area, urban (city) area, or suburban
Rural ____
Directions: Below are some statements about math. Tell us how true or false each
statement is for you by circling the letter that best describes you.
Definitely
False
Mostly
False
A little
bit False
A little
bit True
Mostly
True
Definitely
True
2*
77
11
12
13
14
15
16
17
18
78
19
20
21
22
23
24
*Item wording changed from the original Sources of Middle School Mathematics SelfEfficacy Scale (Usher & Pajares, 2009). The original wording for each item is included
below:
#2 - Seeing kids do better than me in math helps me do better in math.
#10 - I got good grades on my last report card.
79
References
Arbuckle, J. L. (2007). Amos 16 user's guide. Chicago: SPSS.
Bandura, A. (2006). Guide for constructing self-efficacy scales. In F. Pajares, & T. Urdan
(Eds.), Adolescence and education, Vol. 5: Self-efficacy beliefs of adolescents (pp.
307-337). Greenwich, CT: Information Age Publishing, Inc.
Bandura, A. (1997). Self efficacy: The exercise of control. New York, NY: Freeman.
Betz, N. E. (1978). Prevalence, distribution, and correlates of math anxiety in college
students. Journal of Counseling Psychology, 25, 441-448.
Byrne, B. M. (2009). Structural equation modeling with AMOS: Basic concepts,
applications, and programming 2nd Edition. CRC Press.
Carnegie Foundation. (n.d.). Retrieved August 2011, from The Carnegie classification of
institutions of high education - updated 2010:
http://classifications.carnegiefoundation.org//
Cresswell, J. W. (2008). Educational research: Planning, conducting, and evaluating
quantitative and qualitative research. Upper Saddle River, NJ: Pearson Education,
Inc.
Geary, D. C. (1996). Sexual selection and sex differences in mathematical abilities.
Behavioral and Brain Sciences , 19 (2), 229-284.
Graham, S. (1994). Motivation in African Americans. Review of Educational Research,
64 (1), 55-117.
Hailikari, T., Nevgi, A., & Komulainen, E. (2008). Academic self-beliefs and prior
knowledge as predictors of student achievement in mathematics: a structural model.
Educational Psychology, 28 (1), 59-71.
Hall, M., & Ponton, M. (2005). Mathematics self-efficacy of college freshmen. Journal
of Developmental Education, 28 (3), 26-33.
Hodges, C. B. (2008). Self-efficacy, motivational email, and achievement in an
asynchronous math course. Journal of Computers in Mathematics and Science
Teaching, 27 (3), 265-285.
Hodges, C. B., & Murphy, P. F. (2009). Sources of self-efficacy beliefs of students in a
technology-intensive asynchronous college algebra course. Internet and Higher
Education, 12, 93-97.
80
Huck, S. W. (2000). Reading statistics and research. 3rd edition. New York: Addison
Wesley Longman, Inc.
Lent, R. W., Brown, S. D., Gover, M. R., & Nijjer, S. K. (1996). Cognitive assessment of
the sources of mathematics self-efficacy: A thought-listing analysis. Journal of
Career Assessment, 4 (1), 33-46.
Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1991). Mathematics self-efficacy: Sources
and relation to science-based career choice. Journal of Counseling Psychology, 38
(4), 424-430.
Lent, R. W., Lopez, F. G., Brown, S. D., & Gore, P. A. (1996). Latent structures of the
sources of mathematics self-efficacy. Journal of Vocational Behavior, 49, 292-308.
Luzzo, D. A., Hasper, P., Albert, K. A., Bibby, M. A., & Martinelli Jr., E. A. (1999).
Effects of self-efficacy-enhancing interventions on the math/science self-efficacy
and career interests, goals, and actions of career undecided college students.
Journal of Counseling Psychology, 46 (2), 233-243.
Matsui, T., Matsui, K., & Ohnishi, R. (1990). Mechanisms underlying math self-efficacy
learning of college students. Journal of Vocational Behavior, 37, 225-238.
Michaelides, M. (2008). Emerging themes from early research on self-efficacy beliefs in
school mathematics. Electronic Journal of Research in Educational Psychology, 6
(1), 219-234.
Moody, V. (2004). Sociocultural orientations and the mathematical success of African
American students. The Journal of Educational Research, 97 (3), 135-146.
Muthen, B., & Asparouhov, T. (2010, September). Bayesian SEM: A more flexible
representation of substantive theory.
National Center for Education Statistics. (2011). Integrated Postsecondary Education
Data System: College Data. Retrieved from National Center for Education
Statistics: http://nces.ed.gov/globallocator
Ozyurek, R. (2005). Informative sources of math-related self-efficacy expectations and
their relationship with math-related self-efficacy, interest, and preference.
International Journal of Psychology, 40 (3), 145-156.
Pajares, F. (1997). Current directions in self-efficacy research. In M. Maehr, & P. R.
Pintrich (Eds.), Advances in Motivation and Achievement (pp. 1-49). Greenwich,
CT: JAI Press.
81
82
Wadsworth, L. M., Husman, J., Duggan, M. A., & Pennington, M. N. (2007). Online
mathematics achievement: Effects of learning strategies and self-efficacy. Journal
of Developmental Education, 30 (3), 6-14.
Warner, R. M. (2008). Applied statistics: From bivariate through multivariate techniques.
Thousand Oaks, California: Sage Publications, Inc.
Zeldin, A., & Pajares, F. (2000). Against the odds: Self-efficacy beliefs of women in
mathematical, scientific, and technological careers. American Educational
Research, 37 (1), 215-246.
83
VITA
Averett University
Danville, Virginia
PROFESSIONAL EXPERIENCE
2000 present
1996 2000
1994 1996
1993 1994
84