home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!ACADVM1.UOTTAWA.CA!JJASB
- X-Delivery-Notice: SMTP MAIL FROM does not correspond to sender.
- X-Acknowledge-To: <JJASB@ACADVM1.UOTTAWA.CA>
- Message-ID: <STAT-L%93012608112704@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Tue, 26 Jan 1993 08:10:16 EST
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: Antoine Morin <JJASB@ACADVM1.UOTTAWA.CA>
- Subject: Re: Violated t-test assumptions: how to adjust?
- In-Reply-To: Message of Mon,
- 25 Jan 1993 19:55:06 GMT from <ph@PHYSIOLOGY.OXFORD.AC.UK>
- Lines: 14
-
- Re: alternative to t-test when sample size and variance is unequal.
-
- In my naive approach (Im am not a statistician, have no formal training
- in stats, and so on...) I would use a permutation test to build an empirical
- distribution of the statistic and compare the observed value of the statistic
- for the actual sample to the distribution obtained by permutation.
-
- Now, my question to the gurus reading this: Is this a valid approach?
- Other than independence, what other assumption am I making if I make
- an inference by this procedure?
-
- Antoine Morin
- University of Ottawa
- jjasb.acadvm1.UOttawa.ca Honni soit qui mal y pense!
-