Review of the Use of Statistics in Infection andImmunity

The use of statistics in scientific and medical journals has been subjected to considerable review in recent years.Many journals have published systematic reviews of statistical methods (4-8, 12).These reviews indicate room for improvement.Typically, at least half of the published scientific articles that use statistical methods contain statistical errors.Common errors include failing to document the statistical methods used or using an improper method to test a statistical hypothesis.This study analyzes articles in Infection and Immunity for appropriateness of statistical analysis and reporting.I reviewed all 141 articles from two issues, January 2002 (volume 70, no. 1) and July 2002 (volume 70, no.7); listed the statistical analyses used; and identified errors in analysis and reporting.Errors were defined with respect to the current instructions to authors (2), accepted statistical practice (1, 11), and professional judgment.The statistical errors identified in Infection and Immunity are comparable to those found in similar journals: 54% of the articles reviewed contained errors of analysis (20%), reporting (22%), or both (12%).The most common analysis errors are failure to adjust or account for multiple comparisons (27 studies), reporting a conclusion based on observation without conducting a statistical test (20 studies), and use of statistical tests that assume a normal distribution on data that follow a skewed distribution (at least 11 studies).The most common reporting errors are unlabeled or inappropriate error bars or measures of variability (15 studies) and failure to describe the statistical tests performed (12 studies).These errors are discussed more fully below, with examples and suggestions for improvement.