Abstract
The coding of time in latent curve models has been shown to have important implications in the interpretation of growth parameters. Centering time is often done to improve interpretation but may have consequences for estimated parameters. This article studies the effects of coding and centering time when there is interindividual heterogeneity in time such as when longitudinal responses are dependent on a point of origin that varies between individuals. Using representative examples that differ in their degree of interindividual time heterogeneity, we compare different models based on alternative forms of coding and centering of time to evaluate potential for biased estimates. Recommendations are made for studies marked by heterogeneity in time measures.
Original language | English |
---|---|
Pages (from-to) | 413-433 |
Number of pages | 21 |
Journal | Structural Equation Modeling |
Volume | 15 |
Issue number | 3 |
DOIs | |
State | Published - Jul 2008 |