Timescale mismatch in intensive longitudinal data: Current issues and possible solutions based on dynamic structural equation models.

Abstract

Intensive longitudinal data have been increasingly used to examine dynamic bidirectional relations between variables. However, the problem of timescale mismatch between variables faced by applied researchers remains understudied. Under the dynamic structural equation modeling framework, previous studies used the partial-path model and the average-score model, respectively, to explore the dynamic interaction processes and overall reciprocal effects between variables with mismatched timescales. The present study aimed to evaluate the performance of the existing modeling approaches and the effectiveness of the improved approaches (i.e., the full-path model, the factor model, and the adjusted factor model). Study 1 showed that the full-path model, which considered the cross-lagged effects of all time points of variables with denser timescales, better reflected dynamic interaction processes and time-specific effects between variables than the partial-path model. Study 2-1 found that the estimates of autoregressive and cross-lagged effects between timescale mismatched variables were biased in the average-score model, but accurate in the factor model. Study 2-2 further suggested that when there were regression effects between different time points of variables with denser timescales, the adjusted factor model obtained less bias than the factor model, yet the difference is negligible when the regression effects are small. Study 3 used empirical data with timescale mismatched variables to illustrate the differences of all modeling approaches. This study identified the important problem of timescale mismatch in intensive longitudinal data and its possible solutions, providing methodological guidance and valuable insights for data collection and analysis of variables with mismatched timescales. (PsycInfo Database Record (c) 2025 APA, all rights reserved).