home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!gatech!paladin.american.edu!auvm!BINGVMB.BITNET!LJONES
- Message-ID: <SAS-L%93010709332172@UGA.CC.UGA.EDU>
- Newsgroups: bit.listserv.sas-l
- Date: Thu, 7 Jan 1993 08:53:32 ECT
- Reply-To: Lary Jones <LJONES@BINGVMB.BITNET>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: Lary Jones <LJONES@BINGVMB.BITNET>
- Subject: Re: t
- In-Reply-To: Your message of Wed, 6 Jan 1993 22:11:42 EST
- Lines: 23
-
- CONTENT: Additional comments on subject selection
- SUMMARY: Be aware of experimental history effects
- REL/PLTF: All
- E-ADDR: LJones@BingVMB.Bitnet
- NAME: Lary Jones, Computing Services, Binghamton Univ.
- PHONE: 607-777-2879
- ----------------------------------------------------------------------
- Thanks to William Gore for adding randomization to the proposed
- solution for selecting a subset of cases. I seem to have had the same
- reaction to the first two postings that he did.
-
- I have learned to be aware of possible experimental history effects
- (i.e., something happening across the life of the experiment which has
- a systematic effect). Additionally, the number of cases in a condition
- may relate to a naturally occuring frequency or population
- restriction. One should explore the data (as in EDA) _in its entirety_
- for associations of variance, extreme values, etc. with sample size,
- before modifying the sample to be analyzed. In any case, if one
- chooses to analyze a balanced set of cases, a random subset of the
- more populated conditions certainly will be a more representative
- choice.
-
- Thanks Bill!
-