Many studies have found strong peer correlations for a variety of problem behaviors that begin in adolescence (e.g. substance use). Such correlations are commonly attributed to peer influences, but could also be explained by selective ('assortative') friendship: the tendency for those with similar patterns of behavior to become friends. Here we show how, under certain assumptions, cross-sectional data from pairs of siblings or twins and their peers may be used to resolve the contributions of peer selection and reciprocal peer environmental influences to peer resemblance. We performed power calculations to determine necessary sample sizes for rejecting with 80% power, at the 5% significance level, the hypothesis of only peer selection effects, or only reciprocal peer environmental effects. A false hypothesis of only selective friendship effects was always easier to reject than a false hypothesis of only reciprocal peer environmental influences. Limitations of these simulations, including uncertainty about the most appropriate way to model peer selection, are discussed.