Specifying a Formative Model PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 acumberl posted on Tuesday, November 16, 2004 - 12:04 am
I am trying to specify a formative model with 8 X variables (going into success) and two observed y variables coming out of success (predicted by success). This is how my code is written where F1 is success:

F1 by y* y;
F1 on X1@1 X2 X3 X4 X5 X6 X7 X8;


I was wondering if this is correct because I am still getting standard errors for all X's, Y's, and success in my output. I was under the assumption that for a formative model there would be no errors estimated for the X's and the construct. Is there something to change in the code? Thanks so much!
 Linda K. Muthen posted on Tuesday, November 16, 2004 - 10:22 am
I'm not sure if I understand your question. It sounds like you might be confusing standard errors of parameter estimates with residual variances. If you are talking about the second column marked S.E. in the Mplus output, these are standard errors of the parameter estimates.
 acumberl posted on Tuesday, November 16, 2004 - 10:49 am
Hi Linda. I am talking about the standard errors of the parameter estimates. Am I correct in assuming that those should not be present in the model as I specified it? Or, should we just not have the residual variances only for the X's and construct.
Thank you!
 bmuthen posted on Tuesday, November 16, 2004 - 12:55 pm
You get SEs for the parameters that are estimated, which in your case are for example the f1 on x slopes. The x variances and covariances are not part of the model and therefore do not have SEs. Let me know if this doesn't answer your question.
 acumberl posted on Tuesday, November 16, 2004 - 2:35 pm
Hi Bengt. I see. So we are supposed to get SEs for all estimated pieces of the model the way I have it specified. Is there a place on the output that shows where the variances and covariances do not have SE's? I just want to make sure my code is correct. Also, if I don't set one of the X's to 1.0 I don't get SE's for any of the X's, Y's, or the factor. I wanted to make sure that the first X has to be set to 1 in this type of formative model.
Thank you!
 bmuthen posted on Tuesday, November 16, 2004 - 4:26 pm
The x variances and covariances are not reported in the output - they are simply the sample values. If you don't fix the slope at 1 for f on one of the x's, your model is not identified (the output has an error message saying this) and that is why you don't get SEs in that case.
 acumberl posted on Tuesday, November 16, 2004 - 7:59 pm
Thanks for your response. That's also what I thought-- that the model would not be identified, but I'm not getting that error when I don't fix any X's to 1 and there also are no SE's. The model seems to be identified. I've run it other ways where I get the message that the model isn't identified, but not the way I suggested. Any thoughts?
Thank you!
 Linda K. Muthen posted on Wednesday, November 17, 2004 - 6:05 am
Send the output to support@statmodel.com. Then we can see the exact model and see why you don't get standard errors.
 acumberl posted on Wednesday, November 17, 2004 - 11:17 am
Hi Linda. Thanks so much. I have emailed you my output and would really appreciate your guidance.
Thanks!
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: