What multi-level modeling can teach us about single-level modeling and vice versa: The case of latent transition analysis

Bengt Muthén, PhD
UCLA, Professor Emeritus

ABSTRACT:
This talk discusses modeling connections between multi- and single-level analysis in the contexts of latent trait-state modeling, cross-lagged panel modeling, factor analysis, latent class analysis, and latent transition analysis.  Both continuous and categorical outcomes are considered.  A key point is that single-level, wide-format modeling of longitudinal data is sometimes carried out in a way that is different from what one would do in a two-level context.  The single-level modeling can benefit from two-level thinking.  An example of this is latent transition analysis which has been lacking an important two-level component.  Two-level modeling can also benefit from single-level thinking, an example of which is time series analysis of intensive longitudinal data.