home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!uvaarpa!darwin.sura.net!paladin.american.edu!auvm!NOVAVAX.NOVA.EDU!KABACOFF
- X-Mailer: Elm [version 2.1 PL1]
- Message-ID: <9210122108.AA12221@novavax.nova.edu>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 12 Oct 1992 17:08:00 EDT
- Sender: "STATISTICAL CONSULTING" <STAT-L@MCGILL1.BITNET>
- From: "(Robert I. Kabacoff Ph.D.)" <kabacoff@NOVAVAX.NOVA.EDU>
- Subject: path analysis question
- Lines: 16
-
- Please forgive me if this question is basic. I faculty member came
- to me with the following situation.
-
- She is doing a path analysis (recursive and all variables observed) and
- has found that adding variables to the model increased the goodness of
- fit index (chi square) while decreasing the amount of variance accounted
- for in the criterion variable.
-
- I did not think that this was possible. Am I missing something?
- --
- =============================================================================
- Rob Kabacoff, Ph.D. | Internet: kabacoff@novavax.nova.edu
- Center for Psychological Studies | Phone: (305) 475-7581
- Nova University, College Ave. | "Florida is Florida...it's warm, there are
- Ft. Lauderdale, FL 33314 | oranges, and safe driving is optional."
- =============================================================================
-