draw_parameters() throwing error with SingleGroupClass fitted object

[ddueber]

The following code throws an error:

a <- matrix(rep(1, 20), ncol = 1)
d <- rnorm(20, 0, 1)
MyData <- simdata(a, d, 1000, itemtype = "2PL")
mod1 <- mirt(MyData, 1, SE= TRUE)
draw_parameters(mod1, 10)

The error is

Error in draw_parameters(mod1, 10) : 
  no slot of name "ParObjects" for this object of class "dich"

But when I make it multi-group data, everything is ok:

group <- c(rep("G1", 500), rep("G2", 500))
mod2 <- multipleGroup(MyData, 1, group = group, SE = TRUE, invariance = c("free_means", "free_var", "slopes", "intercepts"))
draw_parameters(mod2, 10)

I recognize that draw_parameters() is typically used with DRF, and therefore on MultiGroupClass objects. However, I am hoping to use Lee, Cai, & Kuhfeld's (2016) "poor person's posterior predictive checking" on SingleGroupClass objects.

[philchalmers]

Thanks, I was aware of the problem. Parts of the package related to graphical methods actually do this already, so the fix should be straightforward.

[ddueber]

Thanks for your quick attention. In the meantime, is there a way for me to create a fake MultiGroupClass object and feed the fitted SingleGroupClasss object into the appropriate slot(s)?

[philchalmers]

Interesting question. I suppose you could add in some dummy data to your original dataset, create a group variable to indicate the original and new response vectors, and fix all the parameters in the dummy data group at their starting values. Or, you could try and do the parameter generation manually by extracting the parameters and VCOV matrix directly yourself.

[philchalmers]

The above code now runs correctly with the dev version, so feel free to install for the fix. Thanks for the report!