Estimating longitudinal Partial Credit Models and accounting for lack of measurement invariance using Generalized Structural Equation Modeling in Stata.
Résumé
Self-reported outcomes are often collected longitudinally in psychological research. Individuals may answer to some questionnaire items at various times and on many occasions, making continuous-time models more appropriate than discrete-time specifications. Moreover, Rasch family models allowing for item-level analysis give the opportunity to explore and correct for the potential lack of measurement invariance that might be detected across groups of individuals and/or over time. In this tutorial paper, we show how to estimate longitudinal Rasch models for categorical outcomes, in continuous time, while accounting for the possible lack of invariance of the measure, using the Generalized Structural Equation Modelling framework from Stata software. We illustrate the methodology with real data from a clinical study which aimed at describing the trajectory of anxiety symptomatology of patients on the waiting list for kidney transplantation. We provide Stata code and example data to let readers apply the statistical method presented in the tutorial.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |