대한치주과학회지에 게재된 논문의 통계방법 평가

Abstract

When consulting with the statistician of biomedical field, there are some difficulties between research member and statistician due to their different range of understanding.In order to escape eliciting wrong conclusion due to the error of applying statistic method in periodontology field, the purpose of this article is to present the correct statistic method by pointing out the statistical errors after analyzing the method of articles that were published in Korean Academy Journal of Periodontology and made statistic processes in them. In this study, 488 science papers which being put in Korean Academy Journal of Periodontology published from 2000 to 2006 was divided according to year.Over the papers which had statistical processing method, statistic method of the year were investigated by check list and the errors of the method were divided into these categories;1) Cases of incorrect application of statistic methods 2) Cases which is difficult to interpretate 3) Cases of incorrect application of parametric and non- parametric statistic method and than made comparison and analysis by statistician. Items that showed errors in applying statistic method were total 21 items above all articles and the case that did not apply Post-hoc analysis after ANOVA processing had 22 items(7.1%), the most common case, and the cases that applied Student t-test as the Post-hoc analysis after ANOVA processing, the parametric statistics, when needed to apply Mann-Whitney U analysis(verification) through Bonfferoni correction as the Post-hoc analysis after the procedure of Kruskal-Wallis test, the parametric statistics, includes 20 articles(6.5%) and the case that applied ANOVA when needed to apply Repeated measure ANOVA were 14 articles(4.5%). In the 308 articles that applied statistic methods, 59 articles(50.0%), the largest portion of all, were included the case that applied incorrect parametric statistic method when needed to apply parametric statistics and 38 articles(32.2%) were included in the case that applied incorrect parametric statistic method when needed to apply non-parametric statistics, 14 articles(11.9%) were included in the case of applying incorrect non- parametric statistic method when needed to apply non-parametric statistics, 2 articles(1.7%) were included in the case of applying incorrect non-parametric statistic method when needed to apply parametric statistics and 5 articles(4.2%) were hard to interpretate.