Could not invert information matrix using the mixedmirt()

[陈冠宇]

Hi Phil,

 

First of all,thank you for providing a powerful psychometrics package.

I'm trying to use mixedmirt() to build a multilevel testlet IRT model. This model was proposed by Beretvas and Walker (2012), and was used to detect DIF and differential testlet functioning:

  

In short, uop as ability parameters, β10 as item 1's difficulty (i.e.  item easiness because of dummy code),  βt1o is the testlet 1's easiness,  ut1p is the testlet's residual (i.e person-specific testlet random effect).

I used the data from PISA 2012, and picked 3 testlets from reading. The data is unidimensional (according to the ratio of eigenvalues). Although mixedmirt() estimated the fixed effects, but it couldn't get the SE, the warning informaion was: Could not invert information matrix; may not be identified.

I know maybe I make a mistake in syntaxes.Would you like to offer my some advice about how to identified this model? Or, would you like to tell how to get the SE ? I'm looking forward to your reply.

Best,

Guanyu.

Attached is my syntaxes and part of results:

mixedmirt(data = subset_data, covdata = finaldata_covdata, model = model, 

    fixed = ~0 + items + items:gender + itemorder + itemorder:gender, 

    random = ~1 | itemorder:id, itemtype = "Rasch", itemdesign = itemdesign, 

    SE = TRUE)

--------------

FIXED EFFECTS:

                         Estimate Std.Error z.value

itemordertestlet2          -0.187        NA      NA

itemordertestlet3           0.270        NA      NA

items1:gender              -0.136        NA      NA

items2:gender               0.163        NA      NA

items3:gender               0.354        NA      NA

items4:gender               0.226        NA      NA

items5:gender               0.110        NA      NA

items6:gender              -0.174        NA      NA

items7:gender               0.077        NA      NA

items8:gender               0.083        NA      NA

items9:gender              -0.006        NA      NA

items10:gender              0.249        NA      NA

items11:gender              0.175        NA      NA

gender:itemordertestlet2    0.322        NA      NA

gender:itemordertestlet3    0.418        NA      NA

---

Distinguishing Differential Testlet Functioning From Differential ...