Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: factor analysis/gllamm

From   Sophia Rabe-Hesketh <[email protected]>
To   [email protected]
Subject   Re: st: factor analysis/gllamm
Date   Fri, 11 Apr 2003 12:48:45 +0100

Dear Heidi Gjertsen,

you can easily do confirmatory factor analysis
with mixed responses in gllamm. You can also 
obtain factor scores using the prediction command:

gllapred score, fac

However, I would not recommend adopting the
two-stage approach you suggest, i.e. first
deriving factor scores and then using them
as an explanatory variable in a regression.
Problems with this approach have recently
been pointed out by 

Skrondal and Laake (2001). Regression among
factor scores. Psychometrika 66, 563-576.

Instead, I would use a structural equation
model in which the factor is an explanatory
variable in the regression model of interest.

Please email me personally if you need help
with setting up the models.

The latest version of gllamm is available

Best regards,


At 12:30 PM 4/10/2003 -0400, you wrote:
>I have seen old posts on this topic, but I am still not entirely clear on 
>how to proceed with my estimation. I was planning on using confirmatory 
>factor analysis (with one factor) to reduce my data and create a new 
>explanatory variable to be used in a regression. I believe that my 
>variables can be considered as indicators of an underlying latent variable. 
>The issue is that my data are a combination of binary, likerd scale, and 
>continuous variables. As I understand it, the regular factor command in 
>stata is not appropriate in this case, as it assumes the data are 
>continuous and cardinal. I have been reading about the gllamm procedures 
>and they seem appropriate, but I am not sure of what I need to specify in 
>the commands (e.g. distributions) and how to create a new variable based on 
>my results, i.e. is there something similar to the 'score" command after 
>using 'factor"?
>Any help on this subject would be greatly appreciated.
>Thank you very much.
>Heidi Gjertsen 
>*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index