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

Statistical reporting in clinical trials often lacks consistency. Prior research has shown that p-values are frequently misinterpreted or misused. It was already known that effect sizes and confidence intervals are underreported in many studies. No prior work had resolved the relationship between statistical collaboration and p-value reporting. This gap motivated an investigation into how formal statistical collaboration might influence reporting practices. The uncertainty around this relationship led to a study of rheumatology trials. Researchers wanted to determine if acknowledged statistical help affects p-value usage. The absence of clear evidence on this topic created a need for further analysis.

The aim was to assess the impact of formal statistical collaboration on p-value reporting. The specific problem was the widespread misuse of p-values in clinical research. The motivation was to determine if acknowledged statistical help improves reporting standards. Researchers wanted to test whether collaboration with statisticians leads to better practices. The study focused on rheumatology trials from 2015 to 2016. The goal was to compare two groups of studies: those with and without statistical collaboration. The researchers proposed that collaboration could influence the use of exact p-values. The hypothesis was that formal statistical help correlates with improved reporting.

The study analyzed randomized controlled trials published in four rheumatology journals. Researchers identified studies with formal statistical collaboration through coauthorship or acknowledgments. They categorized articles into two groups based on statistical collaboration. The methods included checking for effect sizes, confidence intervals, and exact p-values. They also examined whether p-values were omitted in baseline data tables. The time frame was 2015 to 2016 to ensure recent data. The sample size was 133 articles, with 28 in the collaboration group. The analysis compared reporting practices between the two groups.

Only 21% of the 133 articles had formal statistical collaboration. In this group, 96% reported effect sizes compared to 71% in the other group. Exact p-values were reported in 88% of the collaboration group versus 69% in the other. The difference in effect size reporting was statistically significant (p=0.01). The difference in exact p-value reporting was marginally significant (p=0.08). The study found no mention of p-values in baseline data for both groups. Confidence intervals were more common in the collaboration group. The results suggest a link between collaboration and improved reporting practices.

The authors proposed that formal statistical help may improve p-value reporting. They suggested that acknowledged collaboration could lead to better reporting standards. The study found a correlation between collaboration and increased use of exact p-values. The results may support the idea that collaboration influences reporting practices. The authors did not claim causation but noted a possible association. They suggested further studies to confirm the findings. The study did not propose new statistical methods or guidelines. The authors emphasized the value of formal statistical collaboration in research.