Data, coding scheme and analysis code for: Brief Dark Triad Measures Distort the Difference Between Psychopathy and Machiavellianism, and a Widely Cited Meta-Analytic Cell Cannot Adjudicate It

This deposit holds the extraction, the coding scheme and the analysis code that produce every number reported in the associated article. It is arranged so that a reader can check any single value against the printed table it came from, or rerun the whole analysis and get the article back. The article is a reanalysis of published aggregate data from four meta-analyses of the Dark Triad. No new participants were recruited and no new data were collected. The estimand is the signed difference between psychopathy's and Machiavellianism's correlation with the same outcome, estimated separately by the measurement instrument the primary studies administered, with sampling covariances derived from each source review's own reported trait intercorrelation. CONTENTS 01_manuscript. The manuscript prepared for masked review, its LaTeX source and bibliography, both figures as vector PDF, and the coding scheme. The coding scheme states the sign convention, how each standard error was derived from the published interval, the Olkin-Finn covariance and the trait intercorrelation used for each review, how negative signs were recovered from the symbol fonts in which the sources set them, the two typographical errors repaired in the source tables, and the sample classification rules. 02_extraction. Every extracted cell, with source, outcome category, trait, k, n, the point estimate, the published interval and the derived standard error; the same broken down by measurement instrument for all three traits; all 186 studies in Table 1 of O'Boyle et al. (2012) with the sample description exactly as printed beside the class assigned to it, the constructs measured and the work outcomes contributed; one row per analysed cell carrying the naive and covariance-corrected standard errors; and the counting behind the composition claim, including the distinction between the sum of two study lists and the number of distinct studies. 03_results. The four tables exactly as printed in the article, and the full console output of the analysis. 04_data_json. The working data the code reads and writes, including the per-category coding notes and the two recorded source typographical errors. 05_code. The analysis in the order the build script runs it, from PDF text extraction through the random-effects models to the figures and the manuscript, followed by a verification pass. REPRODUCING Running the build script runs the whole chain and then reads the finished PDF back and recomputes every reported quantity from the extraction rather than from the file the manuscript was built from. It exits non-zero if any reported number, any journal limit, or the masking is wrong. The verification script can also be run alone against a manuscript that has already been built. Requires Python 3 with PyMuPDF and Matplotlib, and XeLaTeX with Biber, the apa7 class, unicode-math, biblatex-apa and TeX Gyre Termes Math. WHAT IS NOT INCLUDED The four source meta-analyses are copyrighted and are not redistributed. Each is identified in the article's reference list by a full APA 7 reference and a DOI and must be obtained from the publisher. The extraction code reads those PDFs from a data directory; supply your own copies to rerun the extraction stages. Everything downstream of the extraction is included, including the extracted values themselves, so the results can be checked and recomputed without them. A LIMITATION STATED PLAINLY The sample classification of the 186 studies was performed by one coder. There is no second coder and no inter-rater reliability statistic for it. It is published here one row per study, with the sample description exactly as printed in the source table beside the class assigned to it, so that any reader can check every assignment and recompute the proportions under a different classification. INTEGRITY A SHA256 manifest covers every file in the archive.