Skip to main content
  1. Guides/

An Introduction to Forensic Metascience

 Author
Author
Ian Hussey
Developer of the tides and recalc packages

An Introduction to Forensic Metascience
#

An Introduction to Forensic Metascience is a free online book by James Heathers (2025) that provides a comprehensive introduction to forensic meta-science techniques — methods for checking the consistency and credibility of scientific reports.

The book is written in a literate programming style, with all analyses implemented in R code that readers can run themselves. It covers techniques ranging from simple consistency checks to more sophisticated statistical tests, along with visual analysis methods for detecting image manipulation.

Citation
#

Heathers, J. (2025). An Introduction to Forensic Metascience. doi: 10.5281/zenodo.14871843

Topics covered
#

The book walks through a progression of techniques for evaluating the consistency of reported results in scientific papers:

TopicDescription
Basic consistency checksVerifying that percentages sum correctly, internal consistency of reported numbers
GRIMGranularity-Related Inconsistency of Means — determines whether a reported mean is mathematically possible given the sample size and integer data
GRIMMERExtension of GRIM to standard deviations, based on the principle that (degrees of freedom x variance) + (n x mean squared) must be a whole number
SPRITETechnique for reconstructing possible distributions consistent with reported summary statistics
Image forensicsVisual analysis techniques for detecting manipulated images, duplicated Western blots, and other figure anomalies
TriageDeciding which papers warrant deeper investigation
AI in forensic meta-scienceThe potential role and limitations of automated analysis

Who is it for?
#

The book is intended for anyone interested in evaluating the credibility of scientific research, including researchers, research integrity officers, ethics committees, journalists, and aspiring forensic meta-analysts. No prior expertise in forensic meta-science is required, though some familiarity with R is helpful for running the code examples.

Resources
#

Related