Many topics in aging research address questions about group differences in prediction. Such questions can be viewed in terms of interaction or moderator effects, and use of appropriate methods to test these hypotheses are necessary to arrive at accurate conclusions about age differences. This article discusses the conceptual, methodological, and statistical problems in one approach to investigating moderator hypotheses. The subgroup regression approach, in which separate regression analyses are conducted in two or more groups, is widely used in aging research to examine group differences in prediction, but the approach can lead to erroneous conclusions. The moderated regression approach, involving the test of a product term, is described and recommended as an a alternative approach. The question of whether social support has greater beneficial effects for younger or older adults is investigated in a study of 287 recently-bereaved adults ranging in age from 20 to 91. Using the subgroup approach, findings indicated that social support significantly predicted depressive symptoms among younger adults and did not significantly predict depressive symptoms among older adults. The moderated regression analysis, however, indicated no significant age differences in the effect of social support. These results clearly illustrate that the analysis strategy researchers choose may have important bearing on theory in aging research such as conclusions regarding the role of social support across the life span. (Contains 1 figure and 2 tables.)