Psychological Inflexibility and Thoughts of Death or Self-Harm Among Adults With Clinically Elevated Social Anxiety: Preregistration Support Dataset and Analysis Code

This archive contains the frozen deidentified candidate-sample dataset, codebook, sample-flow documentation, provenance/ethics information, and locked R analysis script for a preregistered secondary analysis examining whether AAQ-II-assessed psychological inflexibility is associated with greater frequency of thoughts of death or self-harm (PHQ-9 item 9) beyond concurrent depressive symptom severity (PHQ-8) among adults with clinically elevated social-anxiety symptoms (LSAS-SR ≥60). The candidate sample contains N = 423 participants reconstructed from an original N = 643 anonymous online survey dataset. Eleven multivariate Mahalanobis outliers excluded specifically for the previously published clustering analysis are restored and flagged. The expected primary complete-case sample is N = 411. This secondary analysis is distinct from the previously published person-centered/profile study and does not reproduce, extend, or reinterpret its cluster/profile results. Preliminary feasibility analyses were conducted before preregistration, including inspection of the focal association. The accompanying preregistration therefore transparently states that the analysis is not hypothesis-blind. The purpose of preregistration is to lock the definitive analysis specification, assumption-handling rules, and sensitivity analyses before those analyses are conducted. Version 1.1.0 adds the definitive preregistered analysis outputs generated in R 4.6.1, together with a documented technical correction to the prespecified proportional-odds assumption tests. The frozen dataset, preregistered analysis plan, and original preregistered R script from version 1.0.0 are unchanged. The technical correction addresses an implementation issue in which ordinal::nominal_test() returned missing test statistics because model calls referred to helper-function-local objects. Identical preregistered models were refitted explicitly solely to execute the prespecified proportional-odds tests. No scientific analysis specification, sample definition, variable, scaling constant, exclusion rule, or inferential criterion was changed.