A multi-method latent state-trait model for structurally different and interchangeable methods

Authors

Christian Geiser, Utah State UniversityFollow
Tobias Koch, Leuphana University LüneburgFollow
Jana Holtmann, Freie Universität Berlin
Michael Eid, Freie Universität BerlinFollow

Document Type

Article

Journal/Book Title/Conference

Psychometrika

Volume

82

Issue

1

Publisher

Psychometric Society

Publication Date

11-30-2016

First Page

17

Last Page

47

Abstract

A new multiple indicator multilevel latent state-trait (LST) model for the analysis of multitrait-multimethod-multioccasion (MTMM-MO) data is proposed. The LST-COM model combines current CFA-MTMM modeling approaches of interchangeable and structurally different methods and LST modeling approaches. The model enables researchers to specify construct and method factors on the level of time-stable (trait) as well as time-variable (occasion-specific) latent variables and analyze the convergent and discriminant validity among different rater groups across time. The statistical performance of the model is scrutinized by a simulation study and guidelines for empirical applications are provided.

Recommended Citation

Geiser, Christian; Koch, Tobias; Holtmann, Jana; and Eid, Michael, "A multi-method latent state-trait model for structurally different and interchangeable methods" (2016). Psychology Faculty Publications. Paper 1261.
https://digitalcommons.usu.edu/psych_facpub/1261