Public Review Package for a Japanese LLM Counseling-Evaluation Study

This upload contains an anonymized public review package for a study evaluating large language model (LLM)-generated counseling scripts. The package includes preregistration materials, evaluation prompts, Japanese counseling transcripts with matched English translations, anonymized derived datasets, human and AI evaluator outputs, mixed-model results, final sensitivity analyses, reliability and equivalence analyses, observed and SIMR power analyses, summary tables, manuscript-ready figures, and reproducible analysis scripts. Direct participant and rater identifiers have been removed. Human evaluators are represented by pseudonymous IDs, and derived public-review CSV free-text fields for human comments and evaluator AI responses/reasons are provided in English. The original private participant-application spreadsheet, raw API run folders, API keys, local paths, executed notebooks, and other nonessential or disclosure-sensitive files are not included. The included derived datasets are sufficient to rerun the main crossed random-effects mixed models, final sensitivity analyses, reliability analyses, equivalence tests, power analyses, final summaries, and manuscript figures. Rebuilding the full package from the earliest raw inputs requires private source data that are intentionally excluded.