home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!wupost!uwm.edu!psuvax1!psuvm!auvm!VTVM2.BITNET!GAYLORD
- Message-ID: <STAT-L%92121507582123@VM1.MCGILL.CA>
- Newsgroups: bit.listserv.stat-l
- Date: Tue, 15 Dec 1992 04:50:04 EST
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: Clark Gaylord <GAYLORD@VTVM2.BITNET>
- Subject: Re: software packages
- In-Reply-To: Your message of Thu, 10 Dec 1992 16:37:47 +1200
- Lines: 10
-
- On Thu, 10 Dec 1992 16:37:47 +1200 Bob said:
- >>mood for, from a long list. For baby stat, I rather liked STATA which no one
- >>has mentioned. One of the nicest features it had was creating dummy variables
- >>from character class information (e.g., MALE/FEMALE would automatically be
- >>coded 0,1 into a GENDER variable). . . .
- But in most cases, these should be coded -1/+1 -- obviously there is no
- difference when there are no interactions, but when you compute interaction
- terms directly, the interaction term is more meaningful, imho.
-
- Clark
-