[R] help R IRT simulation

[Doran, Harold]

The simRasch function in MiscPsycho package simulates person abilities using N(mu, sigma^2) and items from U(-x,x), which is not what you're asking for. But, I also am not sure how you would manipulate the other moments. However, item parameters tend not to be normally distributed in operational testing programs, so I am curious why one would do so.

Also, Rasch people really don't like to assume persons abilities are normally distributed. The JML procedures makes no assumptions about the shape of either distribution (persons, items) and MML procedures for Rasch tend to use non-parametric methods so you don't have assume an N(0, sigma^2) for the population distribution.

Of course, you can use N(0, sigma^2) for the population distribution for rasch in an MML procedure (or you can also use N(0,1) if you estimate a common slope), but the rasch community tends to not prefer this approach.

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Stuart Luppescu
Sent: Wednesday, March 03, 2010 12:05 PM
To: Helen Lisman
Cc: r-help at r-project.org
Subject: Re: [R] help R IRT simulation

On Wed, 2010-03-03 at 09:45 -0600, Helen Lisman wrote:
> hello R,
> This is about simulation in psychomtrics in IRT in R. I want to simulate b
> parameters(item difficulty) with moments of fixed values of mean, st.d,
> skewness and kurtosis. Is there any specific IRT package in R with those
> functions to control those moments? I have seen other programs that can
> control mean and st.d but not skewness and kurtosis.

Interesting. I wrote a program in C to simulate item responses according
to the Rasch model, as well as the 2PL and 3PL models. I used either
normal or uniform distribution of the person and item parameters. I'm
assuming that you want to manipulate the skewness and kurtosis of the
generated item and/or person parameters, and then produce simulated item
responses. I don't know how you would manipulate the skewness and
kurtosis. I searched around and rnorm() doesn't include arguments for
skewness and kurtosis. kurtosis() functions exist in e1071 and
fUtilities, but they only return the kurtosis of the input data. Let me
know if you find something.
-- 
Stuart Luppescu <slu at ccsr.uchicago.edu>
University of Chicago

______________________________________________
R-help at r-project.org 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.