How to get IRT "a" and "b" parameters from the probability of correct response?

athens — Thu, 25 Sep 2014 00:39:44 GMT

I am trying to estimate IRT a (item discrimination) and b (item difficulty) parameters from the probability of correct response (=p+ values) conditioned on each of the 8 theta points. Someone told me that I have to use logit function to get the "a" and "b" parameters. So I ran the proc Genmod procedure as shown below, but the result looks not right.

Data temp;

input theta p;

card;

-3.0 .001

-2.5 .015

---

2.5 .950

3.0 .998

;;;;

run;

proc genmod data=temp;

model p=theta / link=logit, dist=poi;

run;

Would you let me know what SAS function I have to use to get the correct "a" and "b" parameters?

Thanks!

Re: How to get IRT "a" and "b" parameters from the probability of correct response?

mohamed_zaki — Thu, 25 Sep 2014 07:41:00 GMT

"The result looks not right?"

How it is not right?

Why do not you use PROC IRT?

or check this paper if it may help

Using SAS for Classical Item Analysis and Option Analysis

topic How to get IRT "a" and "b" parameters from the probability of correct response? in Statistical Procedures

How to get IRT "a" and "b" parameters from the probability of correct response?

Re: How to get IRT "a" and "b" parameters from the probability of correct response?