Acknowledged statistical help and a better use of p values: a proposal.

ABSTRACT

OBJECTIVES: The p-value is commonly misused. We hypothesised that a close cooperation with a statistician would go along with a more proper use of p-values. We considered a close cooperation present, when a statistician was a coauthor, or a formal statistical help was acknowledged in a study report. METHODS: Randomised controlled trials published in 2015-16 in 4 widely read rheumatology journals were searched for a close cooperation with a statistician, the inclusion of effect sizes, confidence intervals, exact rather than relative p-values and the omission of p-values in tables depicting trial entry data. RESULTS: There were only 28/133 (21%) articles in which a formal statistical help was acknowledged (Group I). The rest (Group II) gave no acknowledgement of a close cooperation. Reporting of effect sizes (96% vs. 71%) and exact p-values (88% vs. 69%) were more in Group I (p=0.01, and p=0.08, respectively). CONCLUSIONS: While a formal acknowledgement of a close cooperation was notably infrequent at 21%, this went along with improvement in some aspects of p-value reporting. If substantiated by further studies, we propose that a formally acknowledged statistical help should improve p-value reporting. Like all professionals, statisticians would like their name/office to be formally associated with their good work.