Hello All,
I am using the GLLAMM program in Stata and wanted to know if anyone else is
using the same for fitting ordinal item response models.
I am fitting a simple one parameter as well as a two parameter item
response model. I am interpreting the random effect as the latent variable.
However, I am facing difficulties in introducing other explanatory
variables into the existing item response models. I have looked through the
GLLAMM manual and there arent any similar examples.
The current model I am fitting is :
g(P(y{i}{j}<s)) = k{s} - b{i} - n{j}...Model(1)
where g() is the ordinal probit function. y{i}{j} is the ordinal response
for person j on question/item i. k{s} are category thresholds, b{i} are
item biases and n{j} are the person specific random intercepts, which I
interpret as the latent variables. Model(1) is easy to fit in GLLAMM and is
well documented in Chapter 8 of the GLLAMM manual.
To this i want to incorporate:
n{j}= a*w1{j}+ c*w2{j} + q{j}
That is, the latent variable n{j} is itself a function of person j specific
characteristics, say education w1{j} and income w2{j} and a new latent
variable q{j}. a and c are coefficients of the explanatory variables w1 and
w2.
Thus the final model that i want to fit is :
g(P(y{i}{j}<s)) = k{s} - b{i} - a*w1{j} - c*w2{j} - q{j}...Model(2)
I am using all the items (87 of them) and all the persons (203 of them)
together is estimating this model.
If any does have an idea or suggestion as to how to fit Model(2) in GLLAMM
please do let me know.
Thanks in advance,
Swati
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
