Harmonization of theory, methods, and inferences in addictions research.

Abstract

The articles in the present special section highlight four ways in which our applications of methods, and their harmonization with theory, can hold us back, and each offers an avenue for improvement that brings us closer to our goal of building a cumulative scientific record of the study of addiction. It brings together four articles that are intended to provide new ideas and directions for research on addictive behaviors. It is important for researchers to consider how their study designs, measurements, and statistical tests are specific expressions of the theories they wish to test. Each article illustrates a dimension of the gaps between theory and methods, provides an illustrated example of how to bridge those gaps, and provides easy to follow advice for how to apply these ideas in our own work. By designing for replication (Pearson et al., 2021), considering model-theory harmonization (Littlefield et al., 2021), moving toward plain language interpretation of effects (Halvorson et al., 2021), and thinking of models across levels of analysis (Soyster et al., 2021), we can move toward a more robust, replicable, and impactful science of addictive behaviors.

A Tutorial on Analyzing Ecological Momentary Assessment Data in Psychological Research With Bayesian (Generalized) Mixed-Effects Models

“…Generally, we believe it is more informative to focus on effect sizes and use the posterior distribution and 95% credible interval as measures of uncertainty around the estimate. This encourages readers to think more deeply about the field’s theories and models and in that way build a more robust body of empirical evidence (Halvorson et al, 2022; King & Dora, 2022; McCabe et al, 2022; Tackett et al, 2019).…”

“…This could imply that an important between-persons moderator was missed that could explain for whom we should and should not expect the theoretical prediction to hold true. One way to do this is with the help of the tidybayes package (Kay, 2022). In this case, we have plotted the random per-participant effects of demands on the number of drinks consumed during drinking episodes for 10 of our participants (it is hard to plot slopes for hundreds of participants in a visually clear way; Fig.…”