SPSS factor scores

[Maguin, Eugene]

All: I do not see any attachment when I click on this link

 

I am not exactly sure what nabble is, however the attachments can be found on the website thread:  http://spssx-discussion.1045642.n5.nabble.com/SPSS-factor-scores-td5723392.html.

 

Where are attachments shown?

 

David Disabato:

Please email the attachment directly to me. I’d like to see it.

I did a little search at GMU and see that you are in Psych. I’d guess that in the psych program, there is one or more somebodys that teach quantitative methods, specifically SEM. Forget the statistics department; go see them specifically.

 

This started out with questions about factor scores. But it seems to me that the real problem, the problem that drove you to post here is that your SE model is not working and you need help debugging that. It might be that you’d get better help if you presented that problem AND posted to semnet rather than the spss list, even though amos is an spss product.

 

Regarding the factor correlations. Yes, they make sense. The factor correlations out of amos have been corrected for item reliability and so the correlations are higher but with wider SEs. That said, there’s not much difference between the spss corrs and the amos corrs, which means, I think, that the factor loadings must be very high, probably in the 90s. The other point about the factor correlations is that a pure lag 1 autoregressive model probably isn’t going to fit because such a model asserts that the (WB1,WB3) correlation is accounted for by the product of the (WB1,WB2) and (WB2,WB3) correlations (and so on). So: .86*.87 <> .82. The situation is even worse when you get to the (WB1,WB5) correlation

 

Gene Maguin

[Bruce Weaver]

Gene, scroll right to the bottom of this message in that thread:

http://spssx-discussion.1045642.n5.nabble.com/SPSS-factor-scores-tp5723392p5723418.html

Maguin, Eugene wrote

All: I do not see any attachment when I click on this link

I am not exactly sure what nabble is, however the attachments can be found on the website thread:  http://spssx-discussion.1045642.n5.nabble.com/SPSS-factor-scores-td5723392.html.

Where are attachments shown?

David Disabato:
Please email the attachment directly to me. I’d like to see it.
I did a little search at GMU and see that you are in Psych. I’d guess that in the psych program, there is one or more somebodys that teach quantitative methods, specifically SEM. Forget the statistics department; go see them specifically.

This started out with questions about factor scores. But it seems to me that the real problem, the problem that drove you to post here is that your SE model is not working and you need help debugging that. It might be that you’d get better help if you presented that problem AND posted to semnet rather than the spss list, even though amos is an spss product.

Regarding the factor correlations. Yes, they make sense. The factor correlations out of amos have been corrected for item reliability and so the correlations are higher but with wider SEs. That said, there’s not much difference between the spss corrs and the amos corrs, which means, I think, that the factor loadings must be very high, probably in the 90s. The other point about the factor correlations is that a pure lag 1 autoregressive model probably isn’t going to fit because such a model asserts that the (WB1,WB3) correlation is accounted for by the product of the (WB1,WB2) and (WB2,WB3) correlations (and so on). So: .86*.87 <> .82. The situation is even worse when you get to the (WB1,WB5) correlation

Gene Maguin

[SabatoPsy]

Hi Gene,

Thank you for your input. The attachments should be at the very bottom of the webpage as Bruce mentioned. If you cannot upload them, please let me know the address I can directly email them to you. Yes, I am at George Mason University. I actually came to this listserv, because my current SEM professor did not know why the SPSS factor score correlations were different than the AMOS latent factor correlations. However, I assume this "correction for item reliability" is what is causing the AMOS correlations to have larger magnitudes. I am going to look more into this correction, but that essentially answers my main question, so thank you!

As for the model fit of the cross-lagged autoregressive model, you make a good point. I was taught in class to only do single lags, but I will talk to my professor about how to do multiple lags. I assume another alternative would be to correlate the different well-being factors (or their disturbance terms) at each time point beyond the stability paths (i.e. WB1 with WB3's error, WB2's error with WB4's error, etc.).

-David

On Thu, Dec 5, 2013 at 10:45 AM, Maguin, Eugene [via SPSSX Discussion] <[hidden email]> wrote:

All: I do not see any attachment when I click on this link

 

I am not exactly sure what nabble is, however the attachments can be found on the website thread:  http://spssx-discussion.1045642.n5.nabble.com/SPSS-factor-scores-td5723392.html.

 

Where are attachments shown?

 

David Disabato:

Please email the attachment directly to me. I’d like to see it.

I did a little search at GMU and see that you are in Psych. I’d guess that in the psych program, there is one or more somebodys that teach quantitative methods, specifically SEM. Forget the statistics department; go see them specifically.

 

This started out with questions about factor scores. But it seems to me that the real problem, the problem that drove you to post here is that your SE model is not working and you need help debugging that. It might be that you’d get better help if you presented that problem AND posted to semnet rather than the spss list, even though amos is an spss product.

 

Regarding the factor correlations. Yes, they make sense. The factor correlations out of amos have been corrected for item reliability and so the correlations are higher but with wider SEs. That said, there’s not much difference between the spss corrs and the amos corrs, which means, I think, that the factor loadings must be very high, probably in the 90s. The other point about the factor correlations is that a pure lag 1 autoregressive model probably isn’t going to fit because such a model asserts that the (WB1,WB3) correlation is accounted for by the product of the (WB1,WB2) and (WB2,WB3) correlations (and so on). So: .86*.87 <> .82. The situation is even worse when you get to the (WB1,WB5) correlation

 

Gene Maguin

---

If you reply to this email, your message will be added to the discussion below:

http://spssx-discussion.1045642.n5.nabble.com/SPSS-factor-scores-tp5723392p5723454.html

To unsubscribe from SPSS factor scores, click here.
NAML

--
David J. Disabato

Clinical Psychology Doctoral Student

George Mason University

[hidden email]