Data and analysis code for: Feeding the Tank Without Emptying the Ocean: A Meta-Analysis of Microbial Protein in Aquafeeds

The extracted dataset, the treatment-level dispersion recovered from the primary studies with the page each value was read on, the extraction-validation record, the risk-of-bias assessment, the search log and the complete analysis code for a systematic review and meta-analysis of feeding trials in which whole-cell microbial protein replaced part of the fishmeal in aquafeeds. 169 publications contribute a comparison, of 199 assessed; 790 diets across 48 species, 1979 to 2026. Two primary outcomes: feed conversion ratio, 544 comparisons from 155 publications, and specific growth rate, 504 comparisons from 142 publications, each as the log response ratio against a within-trial control diet. Effect sizes within a comparison series share a control diet, so the covariance between them is computed in closed form rather than assumed, and inference is cluster-robust on the publication. 706 dispersion values were recovered from the growth tables of 98 primary studies. Running code/R/run_all.R from a clean checkout reproduces every estimate, table and figure in the paper from data/meta_analysis_dataset.xlsx. See MANIFEST.json for what each file is, with the counts computed from the files themselves.