One particularly intriguing aspect of IDA is that we are prompted to think more closely about issues not commonly considered in single-study designs. One salient example is sampling. Sampling refers to the mechanism by which a finite group of individual observations is selected from a larger population with the purpose of drawing inferences from the sample back to the population (e.g., Cochran, 1977). Whereas the importance of sampling in clinical psychology is often taken as a given in single-study designs, between-sample heterogeneity due to sampling in multi-study designs is a significant potential threat to the internal validity of an IDA. Most importantly, we do not want to misidentify effects as theoretically meaningful when they are instead artifacts resulting from differences in sampling composition across contributing studies that were not properly modeled. We can consider two well-developed approaches to sampling: model-based and design-based.
