Conference slides on Block-Wise Model Fit for Structural Equation Models with Experience Sampling Data

Structural equation models for experience sampling data have a large amount of manifest variables. However, common fit indices such as chi-squared, CFI, TLI or RMSEA are biased in large models, which will more often lead to the rejection of models which should be acceptable. We propose block-wise fit evaluation as an alternative. The model is first estimated jointly. Then, parts of the variance-covariance matrices are extracted for the manifest variables uniquely associated to each day or other logical block in the data. Block-wise versions of common fit indices are then calculated from these smaller matrices. We show in two simulation studies that (1) block-wise fit can more often identify correctly specified models in a typical experience sampling data scenario compared to global evaluation and (2) block-wise fit can correctly identify misspecified models, except if the misspecification is purely between days. Block-wise fit is not affected by the number of days, that is, the number of manifest variables in the model. Future research and limitations are discussed.

Conference Slides for: Norget, J. & Mayer, A. (2022). Block-Wise Model Fit for Structural Equation Models With Experience Sampling Data. Zeitschrift für Psychologie, 230, 47–59. https://doi.org/10.1027/2151-2604/a000482