Re: Setting values for discrimination parameters

[Phil Chalmers]

Hi Shaun,

See the mirt.model documentation, specifically the START and FIXED elements (you'll need both here to fix the parameters at their respective starting locations). Something like the following example (notice the first item slope is fixed @ the value 1):

mod <- mirt(Science, 'G = 1-4

                                   START = (1, a1, 1.0)

                                   FIXED = (1, a1)')

coef(mod, simplify=TRUE)

$items

                   a1    d1    d2     d3

Comfort 1.000 4.827 2.614 -1.450

Work      1.228 2.927 0.902 -2.269

Future    2.329 5.288 2.239 -1.985

Benefit    1.083 3.339 0.989 -1.683

$means

G 

0 

$cov

  G

G 1

Phil

On Wed, Aug 17, 2016 at 6:58 PM, Shaun Ratcliff <shaun.r...@monash.edu> wrote:

Hello Phil,

I'm a long-time mirt() user, and am hoping to add some complexity to my IRT models that you may be able to help with. 

I am fitting a two group, two dimension, graded IRT model to survey data. I have diffuse normal priors on the discrimination parameters of both dimensions, with the discrimination parameters  of the item that loads strongest on one dimension, set to zero for the other. 

Here is my current model: 

s <- "F1 = 1-28

F2 = 2-29

PRIOR = (1-28, 'a1', norm, 0, 1000), (2-29, 'a2', norm, 0, 1000)"

model <- mirt.model(s)

test.2 <- multipleGroup(all.data2, model, itemtype = 'graded', invariance=c('slopes'))

I would also like to set the discrimination parameters of the items that load strongest onto each dimension to 1 for that dimension, to aid with identifiability (consistent with this paper, for instance). However, I'm not quite sure how to do this. 

Any advice you can provide on this would be greatly appreciated. Thank you in advance (as well as for all of your work on this R package). 

Regards,

Shaun 

Shaun Ratcliff
Lecturer and unit coordinator, Political psychology 
PhD candidate, political science
Consulting, data analysis

School of Social Sciences 
Faculty of Arts
Monash University

Email: shaun.ratcliff@monash.edu