home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!GSUVM1.BITNET!QMDADJX
- Message-ID: <STAT-L%92090500463466@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Sat, 5 Sep 1992 00:45:44 EDT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- From: Alan Jordan <QMDADJX@GSUVM1.BITNET>
- Subject: npar1way
- Lines: 32
-
- To Anyone:
-
- I recently collected some data from a sample of 261 university students
- (Georgia State University). Some of the questions had three levels
- (yes/no/no opinion or did not answer), and some were likert scale
- (strongly disagree to strongly agree). I wanted to test if there were
- a relationship between some demographic variables and the answers to
- individual questions. (ex. is there a difference by race to the question
- "Do you agree with the smoking ban that prohibits smoking inside of the
- university buildings?") Since we had five categories of race and five
- levels to the question, that would make 25 cells for a Chi-square test.
- With 261 subjects we had 19 cells with expected frequencies below five.
- I tried to do a Kruskall-Wallis test using Proc Npar1way in SAS. I
- got the reults of the median scores test, the Savage test, and the Van der
- Waerden test as well. I got some wildly different results from these tests.
- At least one of the four tests found a "significant" difference among each
- question by each demographic variable. Nobody that I talked to was able to
- explain to me what the diffference is between the Kruskall-Wallis, the Savage
- scores test, and the Van Der Waerden test, so I just used the results of the
- Kruskall-Wallis.
- I have three questions:
-
- 1. Why should these tests give wildly different P values such as
- P =< .007 in one test and P=<.17 on another?
-
- 2. What statistical or theoretical criteria should I use to determine
- which test to believe?
-
- 3. Is there any such measure in any of these non-parametric tests that
- would be equivelent to a Tukey or Scheffe post-hoc?
-
- Thanks in advance, Alan Jordan
-