CFA: setting varibles to 1 PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 Eric Green posted on Monday, February 09, 2009 - 8:12 am
Hi,

I have two variables which are highly correlated (.76) in the same factor of a CFA (dtsi_unt and dtsi_fls). When I fix one at 1.00, the other’s estimate is greater than 1.00, and vice-versa. Neither have large residual variances.

I tried fixing the first variable in each model to 1 as follows:

(1)

MODEL:
f1 BY dtsi_unt dtsi_fls dtsi_far dtsi_arg dtsi_agg;

f2 BY dtsi_tlk bsid_wrt dtsi_hth pclc_con pclc_int dtsi_wlk;

f3 BY pclc_amn bsid_sui pclc_dst;

f1@1 f2@1 f3@1;


(2)

MODEL: f1 BY dtsi_unt*1 dtsi_fls*0 dtsi_far*0 dtsi_arg*0 dtsi_agg*0;

f2 BY dtsi_tlk*1 bsid_wrt*0 dtsi_hth*0 pclc_con*0 pclc_int*0 dtsi_wlk*0;

f3 BY pclc_amn*1 bsid_sui*0 pclc_dst*0;

Any thoughts about why this is happening? Thanks!
 Linda K. Muthen posted on Monday, February 09, 2009 - 10:59 am
Factor loadings can be greater than one. In (1) the first factor loading is fixed to one and the factor variances are fixed to one. Only one of these should be done. In (2) all factor loadings and factor variances are free. I don't think this is identified.
 Eric Green posted on Monday, February 09, 2009 - 11:39 am
Thanks for your quick reply. I see what you mean about (1). I think I am clear on the fix. Thanks again!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: