Some of the initial challenges in planning for an IDA study with primary data are related to resources (see sidebar 3), but here we focus on some of the scientific challenges. First, a hypothesis-driven approach is likely to avert many problems. We suspect that the types of questions you may be best able to answer are those that require larger sample sizes, such as distinguishing among predictors of different kinds of low-base rate behavior (i.e., obsessions versus compulsions), determining whether unusual sub-groups of individuals vary in their risk for a clinical outcome (i.e., alcoholic adults with and without comorbid depression), and examining developmental variation in outcomes or predicted associations over time. Each of these types of questions may be difficult to address within any single contributing study, given the relatively lower prevalence of a given behavior, sub-group or age range and more directly tested within IDA given more sample heterogeneity, larger sample sizes, and greater statistical power.
