home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!stanford.edu!bcm!convex!darwin.sura.net!paladin.american.edu!auvm!OREGON.BITNET!JOE
- X-Envelope-to: SAS-L@UGA.BITNET
- X-VMS-To: IN%"SAS-L@UGA"
- MIME-version: 1.0
- Content-transfer-encoding: 7BIT
- Message-ID: <01GR3G5C2PLE8WX01R@OREGON.UOREGON.EDU>
- Newsgroups: bit.listserv.sas-l
- Date: Fri, 13 Nov 1992 08:09:18 -0800
- Reply-To: Joe St Sauver <JOE@OREGON.BITNET>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: Joe St Sauver <JOE@OREGON.BITNET>
- Subject: Re: Conjoint Experts
- Lines: 45
-
- >Date: 12 Nov 1992 22:03:49 -0500 (EST)
- >From: William Kahn <71020.1025@COMPUSERVE.COM>
- >Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- >
- >I have wondered how to do optimal design with comparisons comparisons.
-
- William, I'm getting confused in my old age -- is this the same problem
- that was posed (earlier) yesterday by gnelson@harpo.dev.uga.edu? I.e.,
-
- [>Date: 12 Nov 1992 09:26:31 -0500
- [>From: gnelson@HARPO.DEV.UGA.EDU
- [>Subject: Conjoint Experts
- [>>Greetings! I am assisting a researcher here at the University of Georgia
- [>>who would like to conduct a conjoint study which has 3 factors. It is a 6 x
- [>>5 x 2 design with 60 possible combinations. Obviously one cannot possibly
- [>>digest 60 alternatives and discriminate but only a small amount of them.
-
- The reason I ask is that the earlier posting had made it sound like you
- wanted a fractional factorial design (PROC FACTEX), not an A-optimal or
- D-optimal design (PROC OPTEX).
-
- >We have a random sample of 17 people from an identified population of
- >potential purchasers we can test.
-
- My feeling is that market research subjects are cheap to obtain -- with
- a sample size of 17, I'd feel rather shaky about drawing any inferences,
- unless you are willing to accept quite a modest level of confidence and
- are interested only in fairly large effects. My first suggestion would
- therefore be to increase your sample size.
-
- >Joe S. perhaps suggests that optex may be be useful for this design. I
- >don't see how. It is clear to me that 170 triples rankings should be able
- >to give me information on how to balance product design vs. price, and
- >should give me a measure of between subject and within subject variability.
- >I have no idea how to do this design. Don't know how to analyze either.
-
- Misquote, there. I suggested FACTEX, not OPTEX. Beast of a completely
- different color.
-
- I think the best thing to do at this point would be to sit down with a
- market researcher face-to-face who routinely does mds/conjoint -- there
- are a number of issues that you probably will want to clarify now before
- plunging ahead...
-
- Joe
-