AMOS CFA model respecification

[Donvanwest]

Hi all,
I wonder if anyone could advise me on a few issues relating to how exactly one can go about properly re-specifying a confirmatory factor analysis model?
1) When looking at the modification indices, how large is large enough to make changes? I have a very big model (8 factors and 112 variables- is this too large actually?) and have covaried all errors with MI estimates of larger than 12?Subsequently, the model looks an absolute mess and its still not a good fit.

2)How do you interpret things when the RMSEA suggests the model is a good fit (0.045) but the CFI and NFI are (0.734, 0.515 respectively)

3) How do I select variables to delete. I thought that I was supposed to look under the estimates tab for std residual covariances above 0.4 but this shows only pairs of variables. Does one delete both items?

Thank you very much!

[David Marso]

112 variables? 8 Latent variables?  Why do you think you require on average 14 indicators per construct?
I can't imagine ever finding a good fit!!!
FWIW:  You may wish to reconsider what you are doing after reflecting upon theoretically relevant literature.  Parsimony is a good thing.  Everything but the kitchen sink is not.
------

Donvanwest wrote

Hi all,
I wonder if anyone could advise me on a few issues relating to how exactly one can go about properly re-specifying a confirmatory factor analysis model?
1) When looking at the modification indices, how large is large enough to make changes? I have a very big model (8 factors and 112 variables- is this too large actually?) and have covaried all errors with MI estimates of larger than 12?Subsequently, the model looks an absolute mess and its still not a good fit.

2)How do you interpret things when the RMSEA suggests the model is a good fit (0.045) but the CFI and NFI are (0.734, 0.515 respectively)

3) How do I select variables to delete. I thought that I was supposed to look under the estimates tab for std residual covariances above 0.4 but this shows only pairs of variables. Does one delete both items?

Thank you very much!

[Swank, Paul R]

With that large a model, you are going to have a large number of degrees of freedom. Thus, the chi square test will be quite powerful to detect even small misfit. The fact that the RMSEA is .045 indicates that the chi square test is too sensitive. On the other hand the CFI and NFI compare the model to the null model (as opposed to the chi square and RMSEA, which compare the model to the saturated one) indicates to me that you probably have low correlations among your measured variables. I agree with David. You had best rethink this model.

Paul R. Swank, Ph.D.
Professor, Department of Pediatrics
Medical School
Adjunct Professor, Health Promotions and Behavioral Sciences
School of Public Health
University of Texas Health Science Center at Houston

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Marso
Sent: Sunday, July 22, 2012 10:24 AM
To: [hidden email]
Subject: Re: AMOS CFA model respecification

112 variables? 8 Latent variables?  Why do you think you require on average
14 indicators per construct?
I can't imagine ever finding a good fit!!!
FWIW:  You may wish to reconsider what you are doing after reflecting upon theoretically relevant literature.  Parsimony is a good thing.  Everything but the kitchen sink is not.
------

Donvanwest wrote

>
> Hi all,
> I wonder if anyone could advise me on a few issues relating to how
> exactly one can go about properly re-specifying a confirmatory factor
> analysis model?
> 1) When looking at the modification indices, how large is large enough
> to make changes? I have a very big model (8 factors and 112 variables-
> is this too large actually?) and have covaried all errors with MI
> estimates of larger than 12?Subsequently, the model looks an absolute
> mess and its still not a good fit.
>
> 2)How do you interpret things when the RMSEA suggests the model is a
> good fit (0.045) but the CFI and NFI are (0.734, 0.515 respectively)
>
> 3) How do I select variables to delete. I thought that I was supposed
> to look under the estimates tab for std residual covariances above 0.4
> but this shows only pairs of variables. Does one delete both items?
>
> Thank you very much!
>

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/AMOS-CFA-model-respecification-tp5714360p5714365.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD