Tuyển tập các báo cáo nghiên cứu về y học được đăng trên tạp chí y học Critical Care giúp cho các bạn có thêm kiến thức về y học đề tài: Statistics review 6: Nonparametric methods. | Available online http content 6 6 509 Review Statistics review 6 Nonparametric methods Elise Whitley1 and Jonathan Ball2 Lecturer in Medical Statistics University of Bristol Bristol UK 2Lecturer in Intensive Care Medicine St George s Hospital Medical School London UK Correspondence Editorial Office Critical Care editorial@ Published online 13 September 2002 Critical Care 2002 6 509-513 DOI cc1820 This article is online at http content 6 6 509 2002 BioMed Central Ltd Print ISSN 1364-8535 Online ISSN 1466-609X Abstract The present review introduces nonparametric methods. Three of the more common nonparametric methods are described in detail and the advantages and disadvantages of nonparametric versus parametric methods in general are discussed. Keywords nonparametric methods sign test Wilcoxon signed rank test Wilcoxon rank sum test Many statistical methods require assumptions to be made about the format of the data to be analysed. For example the paired t-test introduced in Statistics review 5 requires that the distribution of the differences be approximately Normal while the unpaired t-test requires an assumption of Normality to hold separately for both sets of observations. Fortunately these assumptions are often valid in clinical data and where they are not true of the raw data it is often possible to apply a suitable transformation. There are situations in which even transformed data may not satisfy the assumptions however and in these cases it may be inappropriate to use traditional parametric methods of analysis. Methods such as the t-test are known as parametric because they require estimation of the parameters that define the underlying distribution of the data in the case of the t-test for instance these parameters are the mean and standard deviation that define the Normal distribution. Nonparametric methods provide an alternative series of statistical methods that require no or very limited assumptions to be made