graded response model with discrimination constraint

[randyp]

I'm trying to specify a two-factor graded response model with equal discriminations within factors that I'm trying to run: 

nc.2Fmodel <- mirt.model('
   F1 = 1, 2, 6, 10, 11, 13, 14, 15, 18
   F2 = 3, 4, 5, 7, 8, 9, 12, 16, 17
   COV = F1*F2
   CONSTRAINT = (1, 2, 6, 10, 11, 13, 14, 15, 18, a1), (3, 4, 5, 7, 8, 9, 12, 16, 17, a1)
')

I'm not quite sure if I've specified the constraints correctly, but after looking at the output of coefficients, there are still more than 2 discrimination parameters, i.e., there should be one per factor: 

> coef(nc.2Fmod, simplify = T)
$items
        a1    a2    d1    d2     d3     d4
nc01 1.660 0.000 2.103 1.074  0.559 -2.346
nc02 2.604 0.000 4.114 2.213  1.831 -1.639
nc03 0.000 1.522 3.786 1.668  1.276 -0.116
nc04 0.000 2.343 4.851 2.272  1.800 -0.247
nc05 0.000 2.378 5.435 2.917  2.103 -0.003
nc06 1.712 0.000 1.914 0.277 -0.516 -3.159
nc07 0.000 1.765 3.328 1.188  0.687 -0.982
nc08 0.000 1.331 2.712 0.698  0.214 -1.493
nc09 0.000 1.857 3.098 0.810  0.393 -1.350
nc10 2.274 0.000 3.147 1.986  0.977 -1.818
nc11 2.449 0.000 4.757 3.022  2.369 -0.606
nc12 0.000 2.016 4.259 2.432  1.653 -0.286
nc13 1.786 0.000 2.109 0.738 -0.310 -2.898
nc14 1.775 0.000 2.300 0.992  0.053 -2.345
nc15 1.384 0.000 2.642 1.354  0.486 -1.829
nc16 0.000 1.457 2.661 0.939  0.564 -1.289
nc17 0.000 1.015 3.141 1.638  1.144 -0.211
nc18 0.826 0.000 1.981 0.931  0.454 -2.019

Any thoughts about if I've specified the constraints correctly? I've also tried changing the constraint to but with no luck, I get the same results: 

CONSTRAINT = (1, 2, 6, 10, 11, 13, 14, 15, 18, a1), (3, 4, 5, 7, 8, 9, 12, 16, 17, a2)

[Phil Chalmers]

The argument is CONSTRAIN, not CONSTRAINT. Dropping the T should make it work.

Phil

--
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.

[Phil Chalmers]

I think you also want the second parameter in CONSTRAIN to be a2, not a1. Otherwise you are imposing equality constraints in location where parameters are not estimated (the locations can be found by using mirt(data, model, ..., pars = 'values')).

Phil

On Sun, Nov 22, 2015 at 6:34 PM, randyp <randy...@gmail.com> wrote:

Ah, thanks for catching that. I've run into this conundrum now: 

> nc.2Fmod <- mirt(nc_items_nm[2:19], nc.2Fmodel, method = 'MHRM')

Error: Constraint applied to fixed parameter(s) 13 19 25 37 43 49 67 91 97  but should only be applied to

                 estimated parameters. Please fix!