Coefficients interpretation in mirt

[ntsi...@gmail.com]

Hi Phil,

I am conducting a multidimentional IRT on an instrument (30 items, five point scale) comprising four latent factors. I am using a simple structure.  How can I interpret the di parameters? Since the number of dimension are above 3 I used the MHRM estimation method.  Does the choice of three proposed estimation methods, QMCEM, MCEM and MHRM, has an effect on the results ?

library(mirt)

model<-mirt.model("

                  INFO=1-9

                  PRESS=10-16

                  INT=17-25

                  ATH=26-30

                  COV=INFO*PRESS*INT*ATH")

model1<-mirt(myfile, model, itemtype = "graded", method = "MHRM")  

[Phil Chalmers]

On Thu, Oct 25, 2018 at 11:10 AM <ntsi...@gmail.com> wrote:

Hi Phil,

I am conducting a multidimentional IRT on an instrument (30 items, five point scale) comprising four latent factors. I am using a simple structure.  How can I interpret the di parameters?

It's an intercept. So, the probability of endorsement when θ₁ = θ₂ = ⋯ = 0. Think standard linear regression models or logistic regression. 

 

Since the number of dimension are above 3 I used the MHRM estimation method.  Does the choice of three proposed estimation methods, QMCEM, MCEM and MHRM, has an effect on the results ?

Theoretically, no, but numerically, yes. Quadrature methods can be poor for integration in multidimensional models, which is why stochastic variants of integration are supported.

 

library(mirt)

model<-mirt.model("

                  INFO=1-9

                  PRESS=10-16

                  INT=17-25

                  ATH=26-30

                  COV=INFO*PRESS*INT*ATH")

model1<-mirt(myfile, model, itemtype = "graded", method = "MHRM")  

--
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.

[ntsi...@gmail.com]

Dear Phil,

Thank you very much for your support.  Can the intercepts be transformed in something more interpretable, like difficulties?  If I understood correctly the answer to my second question is that you are favour of the MHRM estimator.  Finaly, are you planing to teach to a workshop this academic year?

Thanks 

Nikos

[Phil Chalmers]

How is multidimensional difficulty more interpretable than an intercept? Difficulty generally reflects the location where the response function inflects. In multidimensional models, this occur as vector or hyper-vector locations, so as humans we really can't think in those terms too well. Re the MHRM, I prefer this over quadrature methods when those start to fail. It's more of an alternative than a preferred method, since stochastic approximations have several less attractive properties to them. Finally, not this year am I planning any workshops. 

Phil

[ntsi...@gmail.com]

OK. Phil.  Thank you for your response.  You are very clear.  I am sorry to read that you are not going to give a workshop.  Lets hope for the next year.

Nikos