Data, R code, and reproducibility outputs for "Why some hotel water-saving messages are perceived as more effective—and why this advantage fades: Parallel evaluative-cognitive and affective response pathways"

This repository contains the de-identified, data-minimized analysis dataset, English-language R code, variable dictionary, numerical validation benchmarks, and computational outputs supporting a study of hotel water-saving message appeals and their evaluative-cognitive and affective response pathways. The study used a 4 × 2 between-subjects experiment with 814 participants who reported a recent holiday stay in a hotel in Andalusia. Participants were randomly assigned to one of four water-saving message variants inspired by protection motivation theory (self-efficacy, response efficacy, vulnerability, or severity) presented either as a single poster or as multiple posters located at different water-use points. The released CSV contains only variables needed to reproduce the analyses reported in the associated manuscript and its designated sensitivity analyses. It omits direct identifiers, open-text fields, cleaning and replacement indicators, detailed geographic fields, and questionnaire blocks not used in this article. Derived constructs are reconstructed transparently from item-level variables by the R code. The main analysis reproduces planned HC3 treatment contrasts, measurement reliability and descriptives, PME–facet overlap, confirmatory factor analyses, expanded evaluation–emotion CFA models using MLR and ordinal WLSMV sensitivity, formal attenuation across response stages, TOST equivalence tests, multivariate analyses, behavioral sensitivity checks, and manuscript Figures 1–3. A companion pathway script reproduces parallel evaluative-cognitive and affective decompositions, including theoretically relevant affective pathways that can operate in opposite directions. Because proposed pathway variables and outcomes were measured in the same post-treatment session, indirect products are interpreted as associations compatible with mechanisms rather than as temporally identified causal mediation. The definitive analytic sample contains N = 814. Three recorded age values below the prespecified 18–99 descriptive range are represented as missing in the public CSV; this affects age descriptives only. Full archival execution details, package versions, logs, session information, tables, and figures should be added to this description after the successful version-1.0.0 run.