An Empirical Assessment of Inferential Reproducibility of Linear Regression in Health and Biomedical Research Papers

Only 3 out of 14 health research papers using linear regression achieved inferential reproducibility despite similar data and models.

• Why it matters: Variability in modelling decisions and unrecognized assumption violations threaten the reliability of conclusions in health research, impacting clinical decisions and patient care.
• What they did: The study reanalyzed 43 papers with available data and evaluated 20 for computational reproducibility, focusing on assumption violations like normality and independence across three models per paper.
• The result: Frequent violations, especially of independence, often led to inferential failure, wider confidence intervals, and reduced result certainty, underscoring the need for better statistical practices and collaboration.