Dan:
lmer cannot currently be used for the 2PL. As you note, it is straightforward to
estimate the 1PL, but the a-parameters present a current challenge. Doug
mentioned to me the other day he is doing some work on this, so I have copied
him on this reply.
Harold
-----Original Message-----
From: r-help-bounces@stat.math.ethz.ch on behalf of Dan Powers
Sent: Sun 10/1/2006 10:15 PM
To: 'R-Help'
Subject: [R] multilevel factor model in lmer
Hello --
I am curious if lmer can be used to fit a multilevel factor model such as a
two-parameter item response model. The one parameter model is
straightforward. A two-factor model requires a set of factor loadings
multiplying a single random effect. For example, a logit model for the ith
subject responding correctly to the jth item (j=1,..,J) is
logit[p(ij)] = a1*item1(i) + ... + aJ * itemJ(i) +
lambda1*item1(i)*u(i) + ... + lambdaJ*itemJ(i)*u(i)
where the lambdas are factor loadings, with lambda1 fixed to 1.0 and
item1-itemJ are dummy variables for the items.
Thanks,
Dan
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-Daniel A. Powers, Ph.D.
Department of Sociology
University of Texas at Austin
1 University Station A1700
Austin, TX 78712-0118
phone: 512-232-6335
fax: 512-471-1748
dpowers@mail.la.utexas.edu
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
[[alternative HTML version deleted]]