CODEWITHSUNDEEP
rpackagestandard-error
Ester Silva
Ester Silva

Reputation: 680

mirt package - standard error

I am using mirt package to make a 2PL model. I was trying to get the standard error for each estimated parameter.

After using mirt(mydata, 1, itemtype = '2PL',SE=TRUE), i am getting the following results which I am confused which info shows the standard error.

FYI, mydata is sparse and binary. Also, I already read this page, which I did not get my answer.

Results sample for the item 59

$`59`
              a1        d  g  u
par       36.859  -12.266  0  1
CI_2.5  -905.576 -445.648 NA NA
CI_97.5  979.295  421.117 NA NA

Which information shows the standard error? a1, d, g,or u.

Appreciate!

Upvotes: 0

Views: 1245

Answers (1)

tubsly
tubsly

Reputation: 21

to answer your question directly, none of the information in that table is the standard error.

To get standard errors from a fitted mirt object, you need to use the following:

mydata.mirt <- mirt(mydata, model=1, itemtype="2PL", SE=TRUE)
mydata.coefficients <- coef(mydata.mirt, printSE=TRUE)

That should give you a similar list output to what you've pasted in your question, but instead of confidence intervals you should get the SE for each parameter estimate.

Based on the table you've posted, I would expect those SEs to be quite high. It may be worth checking the data and the model to make sure nothing has gone wrong, as the parameter estimates are surprising and the confidence intervals are massive.

Upvotes: 2

Related Questions