Continuous-time structural equation model forests.

What is it about?

Structural equation model (SEM) trees and forests are useful tools for understanding what factors contribute to differences between groups in how they change over time. In past longitudinal research, the estimation of SEM trees and forests has been primarily performed with the R package semtree, which, until recently, only allowed the estimation of discrete-time models. The problem of these models is that they lead to biased estimates when measurement intervals are not evenly spaced, which is frequent in longitudinal studies. As a solution to this problem, recent updates to the semtree package allowed to build SEM trees using continuous-time (CT) models, which do not require fixed time intervals and can handle irregular sampling schemes more effectively. This implementation, however, yielded biased results and involved a computational burden that was prohibitive for realistic models. Only very recent advances in methods and software have made it feasible to use CT-SEM trees and forests in real-world research. In this article, we present a novel implementation of CT-SEM forests that combines the ctsemOMX package for CT modeling, the recursive partitioning infrastructure of the semtree package, and the score-guided covariate testing procedures of the strucchange package. Next, we examine the performance of our approach using simulated data and illustrate it on empirical data from the Survey of Health, Ageing, and Retirement in Europe. Finally, we explore different ways of using the information provided by CT-SEM forests and discuss the benefits and limitations of the approach

Featured Image