What predicts who winds up satisfied with their romantic relationships and who does not? Over the past 40 years, researchers have uncovered a wide array of specific predictors of relationship quality. However, practical research limitations—such as the costs of recruiting couples, participant fatigue, and limited degrees of freedom using traditional statistics methods—has prevented researchers from directly comparing or meaningfully aggregating the predictive power of these variables over time.
We are seeking collaborators for a multilab project designed to overcome these limitations. We plan to compile a large number of dyadic longitudinal datasets with a diverse range of samples and measures. We will then use machine learning (random forests; Breiman, 2001) to predict relationship satisfaction in each dataset, both initially and over time. The random forests method is designed to handle a large number of predictors at once, with the goal of explaining as much variance in the outcome variable as possible.
By comparing results across many datasets, research questions we hope to answer are: 1) How much variance in relationship quality are researchers currently able to predict? 2) Which kinds of psychological measures tend to emerge as the most robust predictors of relationship quality? And 3) how might these results differ across different kinds of relationships and populations?