Date: Tue, 22 May 2001 11:41:26 GMT
Reply-To: Michael Babyak <mbabyak@DUKE.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Michael Babyak <mbabyak@DUKE.EDU>
Organization: Duke University, Durham, NC, USA
Subject: Re: PROC CALIS RAM option interpretation of goodness of fit
indices
Your fit isn't very good by any of the criteria. The GFI-based fits ought
to be at least above .9, some even argue .95. The Chi-square is
significant, indicating large differences between the model-implied and
observed VCV matrix. One also usually hopes Holter's N is at least 200.
You can try specifying alternative models based on modification indicies,
but you must bear in mind that these sorts of modifications are subject to
the same overfitting problems you would see in any p-level based search,
iei, you need to control for the Type I error rate.
Mike Babyak
Petra Stehle <pstehle@tiscalinet.de> wrote:
: Hello, I am not sure how to interpret the goodness of fit indices of CALIS / RAM option.
: If I am not totally wrong the results opf these indices indicate an acceptable fit according to some range, e.g.GFI, AGFI. However, the Probability of Close Fit troubles me. THIS results seems to state that there is NO significant close fit. Could someone have an eye on these indices?!
: Thank you in advance,
: Petra
: /*******************************************************************************/
: /* Results: */
: /* Model1 Model2 */
: /* Fit Criterion 1.8270 1.9881 */
: /* GFI 0.7266 0.7187 */
: /* AGFI 0.3849 0.4727 */
: /* Chi2-Test (Pr>Chi-Square) <.0001 <.0001 */
: /* Probability of Close Fit 0.0000 0.0000 */
: /* Bollen's (1988) Delta2 0.6685 0.6448 */
: /* Hoelter's (1983) Critical N 19 20 */
: /* */
: /*******************************************************************************/
:
: .......the code......
: title 'PROC CALIS path analysis' ;
:
: /* 1 EDU */
: /* 2 GDS */
: /* 3 SE-SIT */
: /* 4 SE-THT */
: /* 5 RMB */
: /* 6 mos_ew */
: /* 7 mos_gh */
: /* 8 mos_ef */
: /* 9 mos_pf */
:
: proc calis data = SASUSER.PATH cov stderr method=ml outram = PATHRAM ;
: var EDU GDS se_sitm se_thm rmbpcl_m
: mos_ew mos_gh mos_ef mos_pf ;
: RAM 1 2 1 X1, /* GDS <- EDU */
: 1 3 1 X2, /* SE-SIT <- EDU */
: 1 4 1 X3, /* SE-THT <- EDU */
: 1 5 1 X4, /* RMB <- EDU */
: 1 6 2 X5, /* MOS_EW <- GDS */
: 1 7 2 X6, /* MOS_GH <- GDS */
: 1 8 2 X7, /* MOS_EF <- GDS */
: 1 9 2 X8, /* MOS_PF <- GDS */
: 1 2 3 X9, /* GDS <- SE-SIT */
: 1 5 3 X10, /* RMB <- SE-SIT */
: 1 2 4 X11, /* GDS <- SE-THT */
: 1 5 4 X12, /* RMB <- SE-THT */
: 1 6 5 X13, /* mos_ew <- RMB */
: 1 7 5 X14, /* mos_gh <- RMB */
: 1 8 5 X15, /* mos_ef <- RMB */
: 1 9 5 X16, /* mos_pf <- RMB */
: 2 1 1 A1, /* error term EDU */
: 2 2 2 B1, /* error term GDS */
: 2 3 3 B2, /* error term SE-SIT */
: 2 4 4 C1, /* error term SE-THT */
: 2 5 5 C2, /* error term RMB */
: 2 6 6 D1, /* error term mos_ew */
: 2 7 7 D2, /* error term mos_gh */
: 2 8 8 D3, /* error term mos_ef */
: 2 9 9 D4; /* error term mos_pf */
: run;
:
--
_________________________________________________________________
Michael A. Babyak, PhD (919) 684-8843 (Voice)
Box 3119 (919) 684-8629 (Fax)
Department of Psychiatry
Duke University Medical Center mbabyak@duke.edu
Durham, NC 27710
_________________________________________________________________
In Flumine Stercoris Noli Undas Facere
_________________________________________________________________
| 
