About ESCIMate

Conservative, assumption-aware statistical consistency verification

What is ESCIMate?

ESCIMate is an open-source tool that parses APA-formatted statistical results from academic papers — PDF, HTML, DOCX, or pasted text — and independently recomputes effect sizes and p-values. It flags discrepancies between reported and computed values, helping researchers, reviewers, and meta-analysts verify statistical consistency.

Unlike simple p-value checkers, ESCIMate computes all plausible effect size variants (d, g, dz, dav, drm, eta², omega², V, phi, R², rank-biserial r, and more), tests both paired and independent designs, and flags ambiguity rather than guessing. It gives researchers the benefit of the doubt while doing the computational legwork.

How It Works

Extract
Parse
Compute
Compare
Verdict

A 12-phase pipeline extracts statistics, normalizes Unicode and PDF artifacts, computes all plausible effect size variants for the detected design, matches against reported values using type-aware comparison, and produces a confidence-weighted verdict with reproducible R code for every result.

What It Checks

t-tests (independent, paired, Welch)
F-tests / ANOVA
Correlations (r)
Chi-square (χ²)
z-tests
Mann-Whitney U
Wilcoxon W
Kruskal-Wallis H
Regression coefficients

Effect sizes: d, g, dz, dav, drm, gz, gav, grm, eta², partial eta², omega², Cohen's f, phi, Cramér's V, R², adjusted R², rank-biserial r, Cliff's delta, epsilon², Kendall's W, standardized beta, and more.

How to Cite

Feldman, G. (2026). effectcheck: Statistical consistency checker for published research results. R package version 0.3.0. https://github.com/giladfeldman/escicheck

Comparison with statcheck

FeaturestatcheckESCIMate
ChecksP-values onlyEffect sizes + p-values
Test typest, F, r, χ², z+ U, W, H, regression
SeverityBinary (pass/fail)6-level with confidence scores
Design handlingAssumes one designTests all plausible variants
Effect size variantsNone20+ (d, g, dz, eta², V, R², ...)
Repro codeNoYes, for every result

Accuracy & Transparency

v0.3.0
Version
132,499 results
Validated on
8,456 articles
From
996+
Automated tests

Known limitations: table-format statistics, sign errors (absolute value comparison by design), some multi-stat sentences (~2% of detections). See GitHub Issues for the full list.

Created by

Gilad Feldman (ORCID)

Related Projects

Scimeto

Citation analysis and academic metrics

MetaESCI — Large-scale meta-science audit of effect size and CI reporting across 7 psychology journals using ESCIMate.

Open Source

MIT licensed. Contributions welcome.