researchzy-meta-analysislisted
Install: claude install-skill Researchzy/Researchzy-academic-skills
# Researchzy Meta-analysis
This skill pools the results of studies the user has already chosen. It turns an extraction table into one effect size per study and comparison, with every conversion declared on its row; pools them with the model a written plan names; reports heterogeneity and a prediction interval; draws forest and funnel plots; runs sensitivity and subgroup analyses; and records every number in `meta-analysis/results.json`. A script then checks that each number in the results text is in that file and that each pooled estimate is reported with its interval, its prediction interval and its heterogeneity.
It does not search or screen (`researchzy-systematic-review`) or judge risk of bias and certainty of evidence (`researchzy-risk-of-bias`), both in plugin `researchzy-evidence`, and it does not audit someone else's meta-analysis (`researchzy-stats-review`, plugin `researchzy-manuscript`), when installed. It does not fit network, individual-participant, Bayesian, diagnostic-accuracy (bivariate) or multivariate models, pool single-arm proportions, rebuild data from survival curves, or compute fail-safe N or p-curves. When the question needs one of these, name the method and the kind of tool it needs, and stop.
## Requirements
Resolve `SKILL_DIR` to the directory containing this `SKILL.md`. Python 3.10 or newer. `effect_sizes.py` and `check_meta.py` need nothing else (scipy, when installed, gives the t quantiles). `meta_pool.py` needs numpy and scipy; the plots nee