get.stats.app

get.stats.app extracts and recomputes reported statistical results from scientific texts by applying the JATSdecoder get.stats() function to identify reported test statistics and verify p-values for standard two-sided and one-sided tests.

Key Features:

• Automated Extraction: Identifies and extracts reported statistical results from scientific text using text-mining approaches implemented in JATSdecoder's get.stats() function.
• Recomputation of P-Values: Recomputes p-values for most standard test results, supporting verification of reported outcomes for two-sided and one-sided tests.
• Versatility Across Formats: Processes multiple input formats, including PDFs from major psychological journals and XML documents from Frontiers in Psychology and articles published by PLoS One.
• Enhanced Detection Rates: Demonstrates superior detection of statistical results compared with statcheck, reproducing manually extracted significant results with high accuracy and improving detection when used alongside statcheck.
• General Applicability: Handles a broad spectrum of textual representations of statistical results found in scientific manuscripts.

Scientific Applications:

• Plausibility and Reliability Checks: Enables verification of the plausibility and reliability of reported statistical findings by extracting statistics and recomputing p-values.
• Consistency and Completeness Verification: Facilitates manual and automated checks of consistency and completeness of reported results within manuscripts.

Methodology:

The method scans large corpora (13,531 PDF files from major psychological journals, 18,744 XML documents from Frontiers in Psychology, and 23,730 PLoS One articles) and evaluates performance against manually coded datasets and by comparison with statcheck.