home *** CD-ROM | disk | FTP | other *** search
/ NetNews Usenet Archive 1992 #18 / NN_1992_18.iso / spool / bit / listserv / statl / 1308 < prev    next >
Encoding:
Text File  |  1992-08-12  |  993 b   |  26 lines
  1. Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
  2. Path: sparky!uunet!paladin.american.edu!auvm!ECUVM1.BITNET!PSWUENSC
  3. Message-ID: <STAT-L%92081215545777@VM1.MCGILL.CA>
  4. Newsgroups: bit.listserv.stat-l
  5. Date:         Wed, 12 Aug 1992 15:45:06 EDT
  6. Sender:       "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
  7. From:         "Karl L. Wuensch" <PSWUENSC@ECUVM1.BITNET>
  8. Subject:      Re: testing dependent variances
  9. In-Reply-To:  Message of Wed, 12 Aug 1992 09:17:00 IST from <WELFARE@ILNCRD>
  10. Lines: 14
  11.  
  12. To test two related samples for homogeneity of variance:
  13.  
  14. Compute the usual F for independent samples, that is, the larger variance
  15. divided by the smaller variance.  Then insert that F in the formula below:
  16.  
  17.     (F - 1) * SQRT(n - 2)
  18. t = -----------------------
  19.     2 * SQRT[F * (1 - r*r)]
  20.  
  21.  
  22. r is the correlation between the two sets of matched scores.  Evaluate the t on
  23. n-2 degrees of freedom, where n is the # of pairs of scores.
  24.  
  25. This technique is from E.J.G. Pitman, Biometrika, 1939, 31: 9-12.
  26.