Comparing Traditional Growth and Time-to-Criterion (T2C) Latent Variable Models

Abstract

This paper compared traditional growth models with the time-to-criterion (T2C) latent variable models to examine if missingness and sample size would have differential effects on relative bias, model convergence rates, and power of growth parameters. Additionally, w the criterion location was also investigated for potential interactions with different levels of missingness and sample size. Results showed both models were comparable in terms of relative bias and power of mean growth estimates as well as predictor effects on growth estimates irrespective of conditions. However, for the T2C model, the time-to-criterion factor and its predictor effect exhibited positive bias in the small sample condition, and this was exacerbated as missingness levels increased as well as when the criterion location was farther out in time. Moreover, the T2C model approach was also associated with increased convergence issues when coupled with increased missing data and a smaller sample size. These results suggest the T2C approach should only be used in circumstances where there is either ample sample size or modest (or no) levels of missingness. Limitations and future research directions are discussed.