Case selection for pathway analysis

Abstract

Introduction Researchers have long recognized that “the cases you choose affect the answers you get” (Geddes 1990). Accordingly, it is critical to select cases carefully and in a transparent manner. This chapter lays out our general approach for selecting cases for pathway analysis. It begins by briefly reviewing the analytic goals of pathway analysis and how they relate to the general criteria for case selection. It then outlines some of the key challenges in applying these criteria and ends with practical advice for implementing these general principles. The goals of pathway analysis and case selection As discussed in the last chapter, pathway analysis ultimately has two goals: (1) to gain insight into the mechanisms that connect some explanatory variable (X1) and some outcome (Y) in specific cases; and (2) to use the insights from these cases to generate hypotheses about mechanisms in the unstudied population of cases that feature the X1/Y relationship. These two goals, in turn, imply several principles for case selection (see Figure 3.1). The first goal of pathway analysis suggests the expected relationship criteria, which means the degree to which individual cases are expected to feature the relationship of interest between X1 and Y given existing theory, empirical knowledge, and large- N studies. It is perhaps obvious, but studying mechanisms that underlie the X1/Y relationship requires identifying cases where the X1 variable is related to the Y, controlling for possible confounds (X2) (Gerring 2007). If the relationship between X1 and Y differs based on the values of X1, then a researcher needs to understand how the relationship depends on the value of X1.