topic Re: PROC IRT True Score Curve in Statistical Procedures

PROC IRT True Score Curve

wadams — Mon, 10 Jun 2019 16:25:41 GMT

Hi Everyone,

Does anyone have good coding resources for outputting a true score curve from PROC IRT results?

Thank you,

Bill

Re: PROC IRT True Score Curve

ballardw — Mon, 10 Jun 2019 17:04:38 GMT

You might provide, as a minimum, the code you are using to create your IRT model so have some idea what your results might look like.

I am not familiar with the term 'True Score Curve' in this context. Can you provide a link to the definition you are using?

Re: PROC IRT True Score Curve

wadams — Mon, 10 Jun 2019 17:26:14 GMT

Hi ballardw - I have not written any syntax for this analysis yet. I am deciding on using PROC IRT or IRTPRO. As an example, if I used PROC IRT to score a five-item examination and, for individuals with an ability of 1.17, the results show that the probability of a correct response to item 1 = 0.99, to item 2 = 0.93, to item 3 = 0.89, to item 4 = 0.78, and to item 5 = 0.74, then I know their expected true score on the original score scale should simply be the sum of these probabilities or 0.99 + 0.93 + 0.89 + 0.78 + 0.74 = 4.33. That is how I am defining "true score" here (though I know it often called a trait score as well).

So, ultimately, I want the true (or trait) score on the y-axis and latent trait on the x-axis (sort of like the attached Figure from the RJ de Ayala IRT textbook). Thank you!

Bill