Now I have the confirmed structure. Here's a description of the Sample 2 analysis codes and outputs file:
Sample 2 Analysis Codes and Outputs
This document contains the complete Mplus input syntax and output for all structural equation models estimated on Sample 2 (N = 3,007) of the study. The file is organized into 14 sections covering seven models, each estimated separately at two timepoints (Time 1 and Time 2), for a total of 14 analyses.
The seven models represent a progression of increasing complexity. Model 1 estimates the base associations among the predictors and math performance; Model 2 adds the interaction between math anxiety and math interest; Model 3 adds the interaction between math anxiety and math confidence; and Model 4 includes both interactions simultaneously. Models 5 through 7 incorporate interactions with student age to test for developmental differences, with Model 7 including all terms together. Each model was estimated once predicting Time 1 math performance and once predicting Time 2 math performance.
All models were fit in Mplus using the MLR estimator with TYPE = COMPLEX to account for the clustering of students within teachers/classrooms (teacher ID as the cluster variable). Models involving latent interactions used TYPE = COMPLEX RANDOM with numerical integration (the XWITH command) to estimate the interaction terms. Analyses were conducted on multiply imputed data (TYPE = IMPUTATION), with results averaged across 20 imputed datasets. Math anxiety, math interest, and math confidence were each modeled as latent factors indicated by two parcels (anxiety by MA1 and MA2; interest by MI1 and MI2; confidence by MC1 and MC2), while age and gender were included as observed variables.
For each analysis, the document includes the full Mplus input syntax (variable definitions, model specification, and analysis options) followed by the complete output, comprising the summary of analysis, sample statistics, model fit information, and model results (unstandardized parameter estimates with standard errors, test statistics, two-tailed p-values, and rates of missing information). These outputs correspond to the structural equation models, effect-size comparisons, and interaction-probing analyses reported in the associated publication.
