The Review of Statistical Methods in Articles of Journal of Korean Ophthalmology Society.
- Author:
Myoung Hee PARK
1
;
Jae Jun LEE
Author Information
1. Department of Ophthalmology, St.Mary's Hospital The Catholic University of Korea College of Medicine, Korea. jjlkorea@hanmail.net
- Publication Type:Original Article
- Keywords:
Inferential statistics;
Ophthalmic journals;
Statistical validity
- MeSH:
Ophthalmology*
- From:Journal of the Korean Ophthalmological Society
2003;44(3):738-743
- CountryRepublic of Korea
- Language:Korean
-
Abstract:
PURPOSE: The purpose of this study is to assess the validity and to review the errors of each category of inferential statistics used in Journal of Korean Ophthalmology Society (JKOS). METHODS: We reviewed 46 original articles of Journal of Korean Ophthalmology Society from January 2000 to March 2000 to assess 4 categories (the methods of statistical analysis, the way how to describe the statistical techniques, the validity of the used statistical techniques, and the interpretation of the results). RESULTS: With the types of statistics, comparison of mean was most commonly used (51.6%), followed by contingency table (16.1%), regression (15.1%), correlation (9.7%), and etc. (7.5%). With the way of describing the statistical techniques, 28 articles contained full description of every used statistical method, 15 had simply the list of statistical techniques, 2 had some methods omitted in description, and 1 was lack of explanation of statistical method. With the validity of the used statistical techniques, the most powerful statistical method was used in 66.7%, less powerful method in 8.6%, and improper method in 24.7% respectively. With the interpretation of the results, 26 articles drew correct conclusions via proper statistical method, 5 had errors in the interpretation of statistical results, 13 directed to incorrect conclusions due to improper statistical techniques, and 2 could not be judged due to omitted description of statistical method. CONCLUSIONS: We found that univariate or bivariate statistics rather than multivariate statistics were more preferred in JKOS, and authors should be more familiar with broad and precise knowledges of statistics.