Lack of statistical rigor in composite materials experimentation and lessons learned from the science of medicine

摘要

A systematic review of one volume within the journal Composite Structures is presented to highlight current practices in planning and interpretation of experiments comparing groups. A brief review of useful statistical methods is presented including power analysis, ANOVA, t-test, p-value, and confidence intervals. The review identified that none of these techniques has been widely adopted by contributors to the journal. Thus, a risk of misinterpretation exists in the published results. Historical context is discussed for the lack of reporting of robust statistics. Cross-disciplinary arguments are presented to encourage the use of robust statistics based on experimental challenges and common practices within the medical research community.