home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!cs.utexas.edu!torn!cunews!nrcnet0!dgbt!ted
- From: ted@dgbt.doc.ca (Ted Grusec)
- Subject: Re: HELP - WITHIN-S ANOVA
- Message-ID: <1992Aug23.062951.3957@dgbt.doc.ca>
- Organization: The Communications Research Centre
- References: <1992Aug20.145549.12978@dgbt.doc.ca> <1992Aug22.214340.10028@ucsu.Colorado.EDU>
- Date: Sun, 23 Aug 92 06:29:51 GMT
- Lines: 44
-
- In article <1992Aug22.214340.10028@ucsu.Colorado.EDU> mcclella@yertle.Colorado.EDU (Gary McClelland) writes:
- >ted@dgbt.doc.ca (Ted Grusec) writes:
- >
- >>My CSS:Statistica (Statsoft) program will not analyze a 2 X 10 X 13
- >>within-subject design. It indicates a 252 dependent variable limit,
- >>while this experiment has 260. All of this despite a recent update
- >>to release 3.1 which I got just days ago,
- >
- >>Is there any commercial package that will do "large" within-subject
- >>designs? Can someone share experiences, make recommendations?
- >
- >No sympathy from me. With such large within-subject designs the key
- >sphericity and homogeneity assumptions are almost surely violated.
- >No one should be doing such large within-s analyses. It is much better
- >to think of the much smaller degree-of-freedom questions you really want
- >to ask of your 260 data values for each subject. Then, any standard
- >regression program can be used to answer those specific questions.
- >For details, see
- >
- >Judd, C.M., & McClelland, G.H. (1989). Data Analysis: A Model Comparison
- > Approach. HBJ (see especially Chapter 14)
- >
- Thanks. But I should point out that, in real world applications of
- experimental design, people sometimes want answers to explicit
- questions that lead you into designs that, whether you like it or not,
- result in violations of theoretical assumptions. Also, some questions
- are pressing - important decisions need to made and there are narrow
- time limits. All of this was true in my case. Having done (and now,
- thanks to |Stat having analyzed) the experiment, it only now is
- obvious which levels of which factors proved uninformative. But there
- is no time to rerun this extremely expensive experiment. For better or
- worse, the decisions, based on the experiment, have been made.
-
- For many years, I taught experimental design and stat. Having long
- since left academia, I am sometimes tempted to offer a course which
- would sprinkle some "real world" constraints into design applications.
- It's something, I'm afraid, many academics are unaware of, as I too
- was unconscious of when I was in that situation.
-
- --
- ==========================================================================
- Ted Grusec Communications Research Centre (Govt. of Canada, DOC)
- 3701 Carling Ave., Ottawa, Ont.K2H 8S2, CANADA Internet: ted@dgbt.doc.ca
- Compuserve: 73607,1576 (613) 998 2762 Fax (613) 993 8657
-