|
|
 |
| Freeing mean and fixing thresholds |
 
|
| Message/Author |
|
|
|
|
I am trying to estimate an ordered probit model where I fix the thresholds to known values and I'd like to free the variance and the mean of the underlying latent variable. Is this possible?
I tried setting parameterization = theta and the following model statements, but I can't figure out how to free the mean of the latent variable to be estimated.
rincom06 ON prestg80 educ age2 age3 age4 female hisp black orace;
rincom06*;
[rincom06$1@9.21];
[rincom06$2@9.90];
[rincom06$3@10.31];
[rincom06$4@10.60];
[rincom06$5@10.82];
[rincom06$6@11.00];
[rincom06$7@11.23];
[rincom06$8@11.41];
[rincom06$9@11.61];
Best,
Shawn |
|
|
|
|
Sorry. I meant to say intercept for the equation predicting the underlying latent variable -- not the mean of the underlying latent variable.
Best,
Shawn |
|
|
|
|
| You can do this by putting a factor behind the DV. You only have to fix 2 thresholds, not all of them. But nothing is really gained by this approach - it is just a re-parameterization. |
|
|
| Back to top |
|
|
