home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!elroy.jpl.nasa.gov!sdd.hp.com!swrinde!news.dell.com!paladin.american.edu!auvm!UTDALLAS.BITNET!WIORKOW
- From: WIORKOW@UTDALLAS.BITNET (John J. Wiorkowski)
- Newsgroups: bit.listserv.stat-l
- Subject: Lotus Regression
- Message-ID: <STAT-L%92072414370459@VM1.MCGILL.CA>
- Date: 24 Jul 92 18:27:33 GMT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- Lines: 13
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
-
-
- You need to be very careful in using the Lotus Regression command. In gen
- eral it works O.K. However in version 2.1(I haven't tried it for version 3.1)
- I tried the classic y = 1 + x + x^2 + x^3 + x^4 + x^5 for x = 1 to 20. The
- built in regression procedure gave the correct coefficients but estimated a ver
- y non-zero residual variance (about 2.0 as I recall) even though there is no re
- sidual error. I called Lotus and they suggested I rearrange the columns to som
- ething like y = 1 + x + x^5 + x^4 + x^3 + x^2 and damn if it didn't work. I
- do not know what numerical algorithm they are using but it obviouly has some i
- nstabilities.
- John J. Wiorkowski
- Professor of Statistics
- University of Texas at Dallas
-