Reviewer 4 statistical sensitivity analyses

Run Stata with the working directory set to this revision_r4_1001 folder after extracting File_S2.zip. The scripts use the unchanged de-identified dataset at ../revision_0928/Dataset_S1.dta, included in the same archive.

1. method_sensitivity.do: unconstrained random-intercept attempts and baseline comparisons. Both free-variance attempts returned e(converged)=0. They are diagnostic attempts, NOT valid effect estimates. See r4_models.log.
2. common_response_grid.do: fixed common-response variance scenarios, 0%, 5%, 10%, 20%, 30%, separately in pooled and strict samples. See r4_response_grid.log and common_response_results.json. The strict 30% model did not converge and is not interpreted.

All outcomes are conditional on the specified equal-unit-loading, orthogonal response factor. Scenario percentages are imposed components of observed standardized item variance, not measured bias rates. No new bootstrap indirect intervals were estimated. Tables S8-S10 in Supplementary Material S2 report the results.

The original model, data, 1,000-replication bootstrap analyses, and earlier verification files remain unchanged. The grid does not replace the primary estimates or the existing strict-sample results.
