Paired versus unpaired significance testing: How improper statistical analysis altered interpretation of posterior surface changes after LASIK

摘要

In a previous series of JCRS articles, Wilhelmus1 4 provided an excellent practical guide to understanding and avoiding common statistical errors in reporting research. One of the articles addressed the common problem of accepting the null hypothesis (ie, concluding that there is no difference between the compared groups) whenever a P value is greater than 0.05 but without establishing that adequate statistical power exists to detect a meaningful difference.3 In the present article, we provide a specific example of another hypothesis-testing error, in which the improper use of statistics altered the conclusion of a published paper regarding posterior surface changes after laser in situ keratomileusis (LASIK).5