A major aim of HLA and disease association studies is to identify the causative HLA factor truly responsible for the association. This is usually difficult due to the pronounced linkage disequilibrium between most HLA determinants. The causative factor must show the strongest association compared to all other factors. Here we describe a simple analysis which can be used to identify which of two factors, say A and B, shows the strongest association. The basic data for the analysis are the entries of the two-by-four table giving the four phenotypic combinations of A and B in patients and controls, respectively. These data are analyzed in various two-by-two tables involving stratification of each of the two factors against the other. A stronger increase of factor A is established if A is significantly associated with the condition both in B-positives and in B-negatives, when this is not true for B in A-positives and A-negatives. Using simulation with control data, it is demonstrated how linkage disequilibrium may influence secondary associations. The analysis may also be used to investigate interaction between HLA factors, but linkage disequilibrium complicates the interpretation in such cases. The method is exemplified using various published data. Finally, some statistical recommendations are given. Thus, we advise that phenotype (marker) frequencies are generally used instead of gene (i.e. allele, or haplotype) frequencies. The importance of correcting p-values, the levels of significance, and the power of Fisher's exact test are discussed.