Related Experiment Videos
Do multiple outcome measures require p-value adjustment?
1Institute of Evidence-Based Chiropractic 6252 Rookery Road, Fort Collins, Colorado 80528, USA. rjf@chiroevidence.com
BMC Medical Research Methodology
|June 19, 2002
Summary
Adjusting p-values in clinical trials with multiple outcomes is debated. While it controls false positives, it may increase false negatives or require larger sample sizes.
Area of Science:
- Biostatistics
- Clinical Trial Design
- Statistical Inference
Background:
- Multiple outcome measures in clinical trials can inflate Type I error rates.
- Unadjusted p-values may lead to misinterpretation of findings due to false significance.
Purpose of the Study:
- To assess the necessity of p-value adjustments in clinical trials with multiple outcome measures.
- To address the increased risk of Type I errors when testing multiple hypotheses.
Main Methods:
- This study evaluates the statistical implications of using multiple outcome measures without p-value adjustment.
- It considers the balance between Type I and Type II errors in hypothesis testing.
Main Results:
- The risk of finding statistically significant results due to chance increases with the number of comparisons.
- P-value adjustments, while reducing Type I errors, can increase Type II errors or necessitate larger sample sizes.
Conclusions:
- Researchers should weigh statistical significance against effect size, study quality, and external evidence.
- Alternative strategies like selecting a primary outcome or using global measures may be preferable to p-value adjustments.