Data and code for "When the standardizer dominates: reporting-format-associated variation in standardized mean differences in biomarker meta-analysis"

This record contains the analysis datasets and code supporting the study "When the standardizer dominates: reporting-format-associated variation in standardized mean differences in biomarker meta-analysis". The data comprise 377 severe-versus-non-severe comparisons extracted from 140 quantitative studies of circulating cytokines in paediatric Mycoplasma pneumoniae pneumonia. For each comparison the record provides the originally reported summary statistics (mean and standard deviation, or median and quartiles), the analysis values and the transformation method used, the within-group coefficients of variation, Hedges g, the log ratio of means, and a provenance table linking every comparison to its source publication and to the specific table or figure from which it was extracted. Also included are all Python and R scripts required to reproduce every table and figure in the article, the output of 55,000 study-clustered simulation runs used to construct the selection-aware benchmark, the CV-harmonisation and conversion-method sensitivity analyses, and the processing script for the external cardiac cfDNA replication dataset. The dataset overlaps with a companion clinical systematic review; the questions, analyses and results reported in the associated methodological article are distinct. Note on interpretation: comparisons flagged by the coefficient-of-variation rule are audit triggers indicating that a reported standardizer warrants verification. They do not constitute a determination that any individual study reported its data incorrectly.