An evaluation of the use of statistical methodology in the Journal of Infectious Diseases

Abstract

One hundred fourteen articles published in the Journal of Infectious Diseases in 1982 were evaluated for the occurrence of eight commonly made statistical errors. Seventy-one percent of Original Articles and 50% of presentations in the Data Forum used statistical methods to analyze results. Almost all of the articles that used statistics contained at least one statistical error. The most common inadequacy, which occurred in 95% of the articles with statistical data, was the statement of a probability value without a complete summary of the statistical results. The most common error was the failure to include a correction for multiple comparisons. These results suggest that a more clearly stated statistical policy, a more explicit set of instructions to authors, and closer editorial attention to statistical methodology, perhaps at the prepublication phase, would improve the validity of articles published in the Journal.