How many questions are responded to, on the 1:7 rating scale? If there's only one question, K-W is certainly appropriate; although even in that case it may be arguable that adjacent intervals on the scale are approximately equal, so that responses can be treated as though they were on an interval scale. If you have as many as 5 questions and you're adding the responses together (to observe a total score or rating between 5 and 35), you can certainly treat the score as interval.
In either case, an ANOVA is appropriate, and you can then carry out your planned comparisons as t-tests at the alpha/2 level for results at the nominal alpha level. If you nonetheless insist on "non-parametric" procedures (or your thesis committee is insisting on them for you ;-), the Kruskal-Wallis test is exactly equivalent to converting the dependent variable to ranks and carrying out ANOVA on the ranks. Then, again, you can perform t-tests on the mean ranks in the same way that you might have done if you chose to treat the data as though it were of interval scale. And, to address your last question, a Mann-Whitney test (aka a Wilcoxon rank-sum test) is exactly equivalent to a t-test on the mean ranks; so yes, you can, but use alpha/2 for your tabled significance level. On 30 Sep 2002, cblack wrote: > I am trying to compare 3 groups of people, each group exposed to a > different situation. Data consists of question responses giving > ratings from 1 to 7. I am only interested in one-tailed comparisons > between group 1 and group 2 and between group 1 and group 3. > Kruskal-Wallis seems to waste power by doing a 2-tailed test and > considering all possible pairs of groups. What else can I use? Is it > appropriate to treat the comparisons as two independent Mann-Whitney > tests? > > Thanks...if possible, please copy responses in e-mail to: > cblack (at) unix (dot) tamu (dot) edu ----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 [Old address: 184 Nashua Road, Bedford, NH 03110 (603) 471-7128] . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
