are significantly more likely to be prospective-longitudinal and to utilize face-to-face interviews [15]. These smaller studies are also more likely to be able to establish temporal order between cause (stress) and effect (depression). In particular, studies of medical illness stressors overcome the problems of variable stressors between subjects and inaccurate retrospective assessment that compromise power in many other GxE studies. However these medical-stressor studies are typically small, and as a result the protocol plan has excluded them. Some studies the protocol includes, no matter how large, must be designated unsuitable for this project if their measures of stress and depression are weak on validity, as is common when data must be collected through the post, telephone, or internet to contain costs of assessing a large sample. When it comes to measuring stress and depression, face-to-face clinical interviews have superior reliability and validity but are more expensive, usually necessitating smaller samples. Again, Culverhouse et al. have emphasized the importance of matching features of a replication analysis as closely as possible to features of the original published study. The original published study used face-to-face clinical interviews. Thus, the protocol plans to include studies that fail to match the design of the original
