WEBVTT 1 00:00:02.066 --> 00:00:08.332 In this video, you learn to perform nonparametric one-way ANOVA using SAS Studio. 2 00:00:08.333 --> 00:00:14.599 Let’s say that we want to analyze the difference in response in a drug trial for rheumatoid arthritis. 3 00:00:14.600 --> 00:00:23.700 Only women participated. Their responses were measured on an ordinal scale of 1 to 5, with 5 being an excellent response. 4 00:00:23.700 --> 00:00:31.200 The distribution of the ordinal outcome isn’t presumed to be normal; therefore, the One-Way ANOVA task shouldn’t be used. 5 00:00:31.200 --> 00:00:39.900 In SAS Studio we begin by clicking Tasks in the panel on the left side of the window to expand that. 6 00:00:39.900 --> 00:00:43.366 The bottom category of tasks is labeled Statistics. 7 00:00:43.366 --> 00:00:50.799 If you do not already see the contents of the Statistics category, double-click Statistics to expand the category. 8 00:00:50.800 --> 00:00:58.933 Next, double-click Nonparametric One-Way ANOVA. 9 00:00:58.933 --> 00:01:03.366 The task menu appears to the right. 10 00:01:03.366 --> 00:01:08.999 I’ll click the Maximize View button to give myself more space. 11 00:01:09.000 --> 00:01:17.866 Before an analysis can be run, I need a data set, a dependent variable, and one classification variable. 12 00:01:17.866 --> 00:01:21.466 The Data tab is already selected here. 13 00:01:21.466 --> 00:01:28.432 Click the icon next to the field for the data set name. 14 00:01:28.433 --> 00:01:36.266 Select a data set for the analysis by double-clicking the name of the library and then double-clicking the name of the data set. 15 00:01:36.266 --> 00:01:42.099 In this demonstration, we will use the arthritis data set from the sasuser library. 16 00:01:42.100 --> 00:01:49.200 This data set is featured in the SAS documentation in the PROC NPAR1WAY chapter. 17 00:01:49.200 --> 00:02:01.100 So first, I expand SASUSER by clicking on the arrow next to it, then I find the arthritis data set and double-click. 18 00:02:01.100 --> 00:02:11.900 To specify the Dependent variable, click the + next to Dependent variable and then click Response. 19 00:02:11.900 --> 00:02:18.966 Notice that the only variables available to me are numeric variables in the SAS data set. 20 00:02:18.966 --> 00:02:27.199 Next, click the + next to Class variable and then double-click Treatment. 21 00:02:27.200 --> 00:02:36.633 This time both character and numeric variables that have not been used already are available to me. 22 00:02:36.633 --> 00:02:46.399 There is one more variable. Freq indicates the number of participants represented by one row of data. 23 00:02:46.400 --> 00:03:00.800 Click the triangle next to Additional Roles and then the + next to Frequency Count, and then select Freq to fill that role. 24 00:03:00.800 --> 00:03:05.766 Notice how the code is updated with each selection that I make. 25 00:03:05.766 --> 00:03:10.799 Next I click on the Options tab to display options for the model. 26 00:03:10.800 --> 00:03:14.800 The tests available to me are grouped by type of difference tested. 27 00:03:14.800 --> 00:03:26.700 There are tests for location only, for scale only, and for both location and scale. 28 00:03:26.700 --> 00:03:34.866 In addition, I can ask for exact statistics. 29 00:03:34.866 --> 00:03:39.066 The Wilcoxon Scores test is selected by default. 30 00:03:39.066 --> 00:03:48.666 This option includes a box plot. I’ll also select the Median Score test, which includes its own plot. 31 00:03:48.666 --> 00:03:53.966 The Median test is most useful for symmetric distributions with long tails. 32 00:03:53.966 --> 00:03:59.332 By default, I calculate p-values using asymptotic methods, but as you’ve seen, 33 00:03:59.333 --> 00:04:11.266 I could have also requested exact p-values through the pull-down menu at the top of the Test options. 34 00:04:11.266 --> 00:04:20.832 I click the running person icon to submit the task code. 35 00:04:20.833 --> 00:04:21.499 Next, I’d like to click on the Code/Results button and then the Results tab under that to expand the Results area. 36 00:04:21.500 --> 00:04:32.566 Next, I’d like to click on the Code/Results button and then the Results tab under that to expand the Results area. 37 00:04:32.566 --> 00:04:37.332 The Wilcoxon Scores table shows the mean rank score for each of the experimental groups, 38 00:04:37.333 --> 00:04:44.833 Active and Placebo, along with their expected values under the null hypothesis of no difference. 39 00:04:44.833 --> 00:04:50.866 The tests include both a normal approximation and a t-approximation. 40 00:04:50.866 --> 00:04:56.632 Both lead to the conclusion that there is a statistically significant difference between the groups. 41 00:04:56.633 --> 00:05:02.033 The analogous ANOVA test is the Kruskal- Wallis test. 42 00:05:02.033 --> 00:05:10.933 With only two groups, it is essentially the same test as the normal approximation test. 43 00:05:10.933 --> 00:05:19.766 The box and whisker plot of the mean scores graphically illustrates the difference between the groups. 44 00:05:19.766 --> 00:05:23.466 Next are the Median test tables and plot. 45 00:05:23.466 --> 00:05:30.932 The score is an indicator of whether the value of Response is above or below the median for all observations. 46 00:05:30.933 --> 00:05:36.699 The value 1 is given to an observation above the median and 0 is given otherwise. 47 00:05:36.700 --> 00:05:42.333 There are adjustments for ties, and so the scores here are not integer values. 48 00:05:42.333 --> 00:05:46.066 The null hypothesis is the same as for the Wilcoxon test. 49 00:05:46.066 --> 00:05:53.466 Once again, the p-value is very small, leading to a rejection of the null hypothesis of no difference. 50 00:05:53.466 --> 00:06:02.099 The median one-way analysis p-value is the same as the two-sample test p-value because there are only two samples in this model. 51 00:06:02.100 --> 00:06:10.833 The frequency plot illustrates the statistics in the tables. 52 00:06:10.833 --> 00:06:17.433 I have shown you how to perform nonparametric one-way ANOVA using SAS Studio.