
Wilcoxon SignedRank with SAS
The Wilcoxon signedrank testA common experiment design is to have a test and control conditions. A two sample ttest would have been a good choice if the test and control groups are independent and follow Normal distribution. If conditions are not met, nonparametric test methods are needed. This section covers one such test, called Wilcoxon signedrank test (equivalent to the MannWhiney Utest) for two samples. The test is preferred when:
Analyzing the data with Wilcoxon signedrank testConsider the following example. For each of the 9 horses, a veterinary anatomist measured the density of nerve cells at specified sites in the intestine. The results for site I (mid region of jejunum) and site II (mesenteric region of jejunum) are measured. Each density value is the average of counts of nerve cells in 5 equal sections of tissue. The question is to test whether the densities of two sites are different. The data is "horse.csv". The data of two sites (of the same horse) are dependent, futher analysis (ignored) also shown that the data are not normally distributed. Therefore, a Wilcoxon signedrank test is recommended. Open the data set from SAS. Or import with the following command. data horse; infile "H:\sas\data\horse.csv" dlm=',' firstobs=2; input house site1 site2; run; A new column "diff" = "site1""site2" needs to be created first, and then proc univariate is run with the following commends. data horse2; set horse; diff= site1site2; run; proc univariate data=horse2; var diff; run; The key output and intepretationNonparametric test to compare respiration between growth and gap area The UNIVARIATE Procedure Variable: diff Tests for Location: Mu0=0 Test Statistic p Value Student's t t 2.823066 Pr > t 0.0224 Sign M 1.5 Pr >= M 0.5078 Signed Rank S 16.5 Pr >= S 0.0547 The SAS procedure "univariate" performs 3 tests, student's t, sign and Wilcoxon signedrank test. Results are shown in the "Tests for Location" table. According to the pvalue of the "(Wilcoxon) Signed Rank", 0.0547. The test is not rejectd at a &alpha level of 0.05. Note that the intepretations are quite different if chosen other test: student's t or sign test. The student's t is not appropriate since the Normal distribtuion assumption is not met. Furthermore, if sign test is used, the question is to test the hypothesis that the difference between two sites are equally positive or negative . Since the sign test has pvalue of 0.5078 and one should not reject the null hypothesis. However, just because the there are equally positives and negatives doesn't mean the mean is zero. For example, the following two samples have same number of positve and negative differences. But the mean of the first sample is obviously bigger than the second, suggesting the population mean difference, &mu_{1} is greater than &mu_{2}. Sample 1 Sample 2 Difference Sign 100 10 90 + 150 25 125 + 60 61 1  46 50 4  
© COPYRIGHT 2010 ALL RIGHTS RESERVED tqin@purdue.edu 