Post-hoc Survey for Power of 119 Negative Results in Korean Journal of Anesthesiology

Lee, Youn Suk; Ko, Hong

Abstract

BACKGROUND: Statistical type II error has seemed to be ignored commonly by medical researchers. To control and present a power value could be helpful to reduce this type of error and to improve a quality of scientific decision making. We performed the post-hoc survey of the power of the negative results in Korean Journal of Anesthesiology (KJA).
METHODS
One Hundred nineteen articles with negative results published in KJA during a year of 1997 were selected. We collected the numbers of the sample size and calculated the power of the given negative result only when applicable. And each author's attitude to negative results was taken by arbitrary criteria.
RESULTS
Median sample size of these negative results was 16 12 (median interquartile range). We can calculate the power only in 43 articles of 119 negative results. Median power is 18.0% (interquartile range 26.0). In thirty six articles (83.8% of 43) the powers are proved to be under 80.0%. And 22 articles (51.2% of 43) have the powers even under 20.0%. We couldn't find any author who included either power or effect size in the article, and there was only one article in which its authors considered their inadequate number of sample size.
CONCLUSIONS
We conclude that authors of KJA tend to ignore statistical type II error. In 119 negative results published in KJA during 1997, the calculated powers were very low and were not reported in the text.