The Role of Time Metric in Mediation Models with Latent Change Scores: An Example from a Longitudinal Alcohol Study

Introduction: Time metric is an important consideration for all longitudinal models because it influences the interpretation of estimates, parameter estimate accuracy, and model convergence in longitudinal models (O’Rourke, Grimm, & MacKinnon, in preparation). Currently, the literature on latent change score (LCS) models does not discuss the importance of time metric. Latent change scores are well-suited to assess change in alcohol studies as they model dynamic change over time (Bell & Britton, 2014; Witkiewitz, 2011). Also, prior research suggests that many alcohol outcomes are mediated by mechanisms of change (Longabaugh, 2007; Kelly, Magill, & Stout, 2009). This study examined how time metric influenced the interpretation and accuracy of estimates in a LCS mediation model for alcohol outcomes using data from a longitudinal alcohol study (Chassin, Pitts, DeLucia, & Todd, 1999; Chassin, Rogosch, & Barrera, 1991).

Methods: Two LCS mediation models with differing time metrics were fit to a longitudinal data set containing information on multiple time metrics. Parameter estimate values and variances of parameter estimates were compared across models, as well as convergence and fit information.

Results: Preliminary results indicated that specificity of time metric influenced convergence, values and variance of parameter estimates, and acceptability of the model itself.

Conclusions: This example extends prior simulation work examining time metric in latent change score models. In summary, it is important to select a time metric that appropriately models the change process of interest when investigating latent change score mediation models.