[Xcalibre 4.1]-Pure logistic model vs Logistic approximation

[walker22k]

I have started using the brand new Xcalibre 4.1 software for Item Response theory and there are a number of elements I am unfamiliar with. The program requires me to state if I am using a pure logistic model or a logistic approximation. I have created a measure of 29 questions and each item is on a scale of 0-6. I understand a logistic approximation is an approximation of the normal curve, but I do not know what a pure logistic model is or which I should be using

Can anyone tell me which is right for me? Much appreciated!

 

[Masteras]

i think what it means, is the link function, does it say probit and logit?

 

[walker22k]

No, it says Model Constant: D=1.0 (pure logistic model)
D=1.7 (logistic approximation)

And I am supposed to choose one or the other. Thanks for the reply!

 

[spunky]

Masteras is exactly right... an intercept of 1.7 is the psychometric improvement of the more widely used pi/sqrt(3) approximation of a probit mode which minimises the discrepancy between the logistic and the normal distributions... whether you choose a logistic over a probit model (or vice-versa) is always subject to debate in the psychometric literature... none is better than the other one, it depends on the assumptions behind what you're trying to model...

 

[Masteras]

I appreciate the help spunky, and I am sure our friend will too

 

[spunky]

hey, it was all you who pointed out the correct difference, i just placed it on a psychometric framework which i guess could help expand a little bit on the answer...