Tuesday, February 1, 2022
2:00 - 4:00 PM (PST) in Berkeley Way West room 1212 and Zoom
Authors: Perman Gochyyev & Mark Wilson
The main focus of this talk is on Lord’s paradox within the latent variable modeling framework. Lord (1967) describes a hypothetical paradox in which two researchers, analyzing the same dataset using different but defensible methods, come to very different conclusions about the effects of an intervention on outcomes. Lord’s paradox has re-emerged in many causal inference settings today around the issue of when it is appropriate to control for baseline status. This talk will discuss two main approaches for analyzing such data: (1) to regress the change from pretest to posttest on the treatment indicator, and (2) to regress posttest on treatment indicator and pretest. The talk will show how these two approaches can yield conflicting results—hence the paradox. We will elaborate on how approaches can be reconciled, depending on the context, and provide examples from the BEAR Center’s ADM project.