[Rasch] GPCM - number of threshold-parameters and number of response-categories

Fischer, Felix Felix.Fischer at charite.de
Thu Mar 10 02:32:10 EST 2011


Hi list,

i want to use R to estimate ability-levels for a given generalized partial credit model. There is a function to do so called irt.ability() from the package MiscPsycho. As argument, i give the observed response-pattern as well as the item-parameters of the model.

Now, here's my problem: when i give 4 threshold-parameters for one item, irt. ability() seems to construct just 4 response categories. But there are 5 response-categories in my data, which was used to estimate the model and this model just has 4 threshold-parameters. These thresholds are those theta-values, where the probability-functions for the response-categories cross.

I would be glad, if someone can give me some insights on this. Are there different types of threshold parameters?

Thanks
Felix

For those, who use R:



library(MiscPsycho)

?irt.ability()

params <- list("3pl" = NULL, "gpcm" = list(a = c(3.4,1), d = list(item1 = c(-1.15,-.40,.35,1.2), item2 = c(0,.5,1, 1.5))))



irt.ability(c(1,3), params, method = "EAP")

irt.ability(c(2,3), params, method = "EAP")

irt.ability(c(3,3), params, method = "EAP")

irt.ability(c(4,3), params, method = "EAP")

irt.ability(c(5,3), params, method = "EAP")  #there seems to be no fifth categorie?




Dipl. Psych. Felix Fischer
Institut für Sozialmedizin, Epidemiologie und Gesundheitsökonomie
Charité - Universitätsmedizin Berlin
Luisenstrasse 57
10098 Berlin

Tel: 030 450 529 104
Fax: 030 450 529 902
http://epidemiologie.charite.de


-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20110309/3253415f/attachment.html 


More information about the Rasch mailing list