Developmental data science: How machine learning can advance theory formation in Developmental Psychology

Abstract

This is a preprint paper, generated from Git Commit # 48d2b17a. This work was funded by a NWO Veni Grant (NWO Grant Number VI.Veni.191G.090). Andreas Brandmaier and Oisín Ryan provided valuable feedback on the first draft of this manuscript.

Using machine learning to identify early predictors of adolescent emotion regulation development

“…It is also noteworthy that some predictors had relatively high predictive importance, but almost no marginal effect (e.g., father's age, drug use, openness to experience). Such predictors likely derive their importance from interactions (Van Lissa, 2022a, 2022b). For instance, drug use might be a vehicle for social status or self‐harm; older fathers might be an asset if they are involved in children's lives but a liability if they prioritize professional success; openness to experience may be an asset if it leads to self‐discovery but a liability if it leads to risk behavior.…”

“…Second, latent class growth analyses were conducted to determine the presence of youth with trajectories that indicated difficulties in emotion regulation (e.g., lower levels or a temporary dip). We used the R‐package tidySEM (Van Lissa, 2022a, 2022b) to estimate and compare one to six class solutions of the aforementioned curvilinear growth curve model; analysis results are displayed in Table S2. To ensure model convergence in small classes, the latent variable covariance matrix was restricted to zero, reflecting the assumption of homogeneity of trajectories, with remaining heterogeneity attributed to error variance.…”

Current perspectives in developmental science: Introduction to the 2023 special issue

“…A final note: there were three additional articles that were meant to be published as part of this special issue, but due to a production error were published in the previous issue. These papers, by Fong et al (2023), Hendry and Scerif (2023) and Veldkamp and Kemner (2023), are just as good as all of the others, and are part of this special issue, at heart.…”

Identifying important individual‐ and country‐level predictors of conspiracy theorizing: A machine learning analysis

“…This inevitably overlooks other possible predictors beyond the scope of these theoretical models (Douglas et al., 2019), and the relative importance of these predictors (Hornsey et al., 2023). Machine learning offers a complementary approach—a data‐driven, exploratory analysis of many candidate predictors at different levels of analysis, which identifies the most important predictors of conspiracy theorizing and holds the potential to reveal overlooked factors at the individual and contextual (country) level, thus inspiring new hypotheses (Van Lissa, 2022a). In the present study, we conducted a machine learning analysis of 115 potential individual‐ and country‐level predictors of conspiracy theorizing in a large international dataset collected during the early weeks of the COVID‐19 pandemic.…”

“…Existing research has largely been confirmatory, relying on theory to identify reliable predictors of conspiracy theorizing. Recently, it has been argued that machine learning analyses can complement existing theory by facilitating the rigorous exploration of large datasets, casting a broader net and identifying potentially overlooked relevant predictors (Van Lissa, 2022a, 2022b). In a recent analysis of this type, Brandenstein (2022) analysed several predictors of conspiracy theorizing based on Douglas et al.’s (2017) framework of epistemic, existential, and social needs.…”