 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: Scale construction and item analysis (teaching materials ?)
Stata has many psychometric models, but they are often not called that.
(1) Generalizability Theory: The random effects models can be fit using -xtmixed-.
(2) IRT: You can fit a Rasch model using -clogit- for conditional ML estimation and -xtlogit- or -xtmelogit- for marginal ML estimation. Phil Ender has a nice example using -clogit- for the famous LSAT data on the web.
The two parameter logistic 2PL (and many other models) can be fit using -gllamm-. You could to set up a 3PL in -gllamm- as composite link model but I wouldn't hold your breath as the 3PL is notoriously difficult to fit without using penalized ML/Bayesian estimation.
Carolyn Anderson's method for pseudolikelihood estimation of IRT models by rest scores could also be done in Stata as all that is required is -logit-, robust standard errors and a bunch of data manipulation. (Sorry I don't have the cite here and am on my handheld while sitting on the train. I will dig it out later.)
There is a Free IRT group (Google for it) that has some IRT material for Stata.
Also, there's a Stata package for Mokken (nonparametric) IRT, I believe it's called -mokken-.
(3) Many agreement statistics can be gotten using the Somer's d package (-somersd- I believe). Many latent class agreement models not discussed in Streiner and Norman but found in the literature could be estimated using -gllamm-.
-----Original Message-----
From: "Jens Lauritsen" <[email protected]>
To: [email protected]
Sent: 12/30/2008 3:40 AM
Subject: st: Scale construction and item analysis (teaching materials ?)
I recently attended a course on evaluation of health measurement scales
where examples were shown based on SPSS. I am about to produce a "short
overview of Stata Commands for Scale validation and Interrater agreement
investigations" for the participants. The course included aspects of Item
Response Models. The short overview will focus on "classic methods" and
mention IR models.
The course was based on the book:
Streiner DL, Norman GR. Health Measurement Scales - A practical guide to
their development and use. Oxford Univ. Press, 2008 (4ed). ISBN
978-0-19-923188-1
For the overview/notes/"cheat sheet" I am aware of the various parts of
the manuals (kappa, alpha, ROC curve ....etc. ) and the chapter in Juul
(Stata for Medical ....) doing this, but some parts seem to be missing, in
particular in relation to implementations in Stata of: fixed and random
variance components from repeated measures Anova (Two-way Anova) (as
described in chapter 9 of the mentioned book) - Socalled G and D studies.
Minimally important change estimation etc.
If someone have materials or lists describing this report to the list or
contact me off-list at jlauritsen (at) health.sdu.dk
Regards and happy new year to all
Jens Lauritsen, MD PhD
University of Southern Denmark and Odense University Hospital
Denmark
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/