Abstract
In the area of pain clinic, routine statistical tests have been performed to compare effective rates between two kinds of treatments. In most of the statistical analyses, only significance levels are evaluated, without referring to statistical power and required sample sizes. Such approaches are inappropriate and may reach misleading conclusions. This paper presented a method to estimate statistical power and required sample sizes in comparing effective rates between two kinds of treatments. Applications were given to clinical reports in which effective rates of two kinds of treatments were compared with routine statistical tests without considering statistical power and required sample sizes. The results of the applications showed that in cases where sample sizes were smaller than required ones, differences of effective rates could not be detected due to the lack of statistical power, and that in cases where sample sizes were larger than required ones, statistical power became large enough to detect almost the same magnitude of differences of effective rates. The present study indicates the necessity of evaluating statistical power and required sample sizes for appropriate comparisons of effective rates.
