Question about Zh-Value and Number of Variables

[Juliane Rutsch]

Dear Phil,

I have two questions about MIRT:

1. Reagarding the Itemfit-Statistics: What does the Zh-Value stand for? Do higher values or lower values indicate a good itemfit?

2. I tried to fit a mulitdimensional IRT-Model with 12 Factors and 121 Items. R gave me an error massage: "Error: Greater than 20000 quadrature points." I don't really know what that means. Is there any possibility to fit a model like that?

Best regards

Juliane

[Phil Chalmers]

Hi Juliane,

On Thu, Nov 12, 2015 at 6:48 AM, Juliane Rutsch <rutsch...@gmail.com> wrote:

Dear Phil,

I have two questions about MIRT:

1. Reagarding the Itemfit-Statistics: What does the Zh-Value stand for? Do higher values or lower values indicate a good itemfit?

Zh is somewhat strange. According to the original authors higher values indicate that an item fits 'too well', while negative values indicate that the item fits poorly. Though in all honesty, I have little faith about Zh as a real measure of misfit.

 

2. I tried to fit a mulitdimensional IRT-Model with 12 Factors and 121 Items. R gave me an error massage: "Error: Greater than 20000 quadrature points." I don't really know what that means. Is there any possibility to fit a model like that?

That's an enormous number of traits, so it might be better to take subsets of items if at all possible and model those instead. Also, large number of traits in IRT are problematic because they become less accurate and more computationally demanding (relates to numerical integration, but just take my word for it). You also could try the MHRM algorithm, but even at that 12 traits is rather large. It might be worth reading my 2012 paper about the package to get a better understanding of the issue. Cheers.

Phil

 

Best regards

Juliane

--
You received this message because you are subscribed to the Google Groups "mirt-package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mirt-package...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.