How to define and test explanations in populations

Abstract

Solving applied social, economic, psychological, health care and public health problems can require an understanding of facts or phenomena related to populations of interest. Therefore, it can be useful to test whether an explanation of a phenomenon holds in a population. However, different definitions for the phrase “explain in a population” lead to different interpretations and methods of testing. In this paper, I present two definitions: The first is based on the number of members in the population that conform to the explanation’s implications; the second is based on the total magnitude of explanation-consistent effects in the population. I show that claims based on either definition can be tested using random coefficient models, but claims based on the second definition can also be tested using the more common, and simpler, population-level regression models. Consequently, this paper provides an understanding of the type of explanatory claims these common methods can test.

Keywords

• 62a01
• 62f03