Most software packages for analyzing polytomous models under the Item Response Theory approach show option characteristic curves as an output. Given that I have the data on those option characteristic curves, how would I calculate the item characteristic curve?
Hi Augustin;
You can use R (with ltm package) for this purpose. And this link may help you how can conduct the analysis. Good luck.
http://www.uwo.ca/fhs/tc/labs/12.PolytomousIRT.pdf
Hello Abdullah,
Thank you for your answer, but after you perform IRT analysis, say, under the graded response model, you can see the option characteristic curves, as in the document you linked (e.g. on page 7). But what if I want to calculate the item characteristic curve (not item information curve), which shows the relation between the ability level (theta) and the probability of an answer? Is that possible? For example, GGUM2004 software, that runs IRT analysis under the generalized graded unfolding model, shows both the option characteristic curves and item characteristic curves. I cannot find such a function in R.
Dear Mutak
If you use GRM then you will extract the Category Characteristic Function for each item which is the same of ICC but for polytomous items. ICCs in general use for ex. to estimate items parameters and the monotonicity of the item curves, the graphs give you good information which help you for better understanding of the tables you may have as an output.
may be we need some more information about what would you exactly need and what for to help you
Regards
Rigth, only to clarify the library in which is the function grm is the library ltm,, which is aprorpriate for doing binary and ordinal (polytomous items). All details for using this library are in : Article ltm : An R Package for Latent Variable Modeling and Item Res...
It is very nice and easy to use, since you can get the parameter estimace and the ICCs curves.
Dear everyone,
Here's an example of what I'd like to do. I've been running some analyses under the Generalized Graded Response Model using GGUM2004 software. After the analysis is done, some of the charts you can see are the category probability functions (CPF) and item characteristic curves (ICC). I'm uploading an example here, in which you can see CPFs and ICC for the same item. Using R's ltm package I wasn't able to find item characteristic curves for polytomous items (GRM or GPCM) but only the category (option) characteristic curves.
Mutak
CPFs is the ICCs for polytomous items, but you have an ICC for each category, in dichotomous items which only have one right option i.e 1 there are only one curve for that 1s option.
you have 7 curves for your 7 points likert items, you can see that this item have somehow a high peaks probability fns and not too much overlapping, so its discrimination parameter would be good.
Yes, but in the second picture there is only one curve and it is the item characteristic curve of the same item whose 7 category probability curves are shown in the first picture.
I've never seen that conceptualization of polytomous ICC used but it looks like it's just a point-weighted sum of the cat/option functions. If you can output the CRFs, use those, otherwise a few formulas in Excel would be sufficient to calculate this all from the item parameters.
Dear everyone,
Thank you for your help. As I've found out, the polytomous ICC is indeed just a point-weighted sum of the category/option functions, like Mr. Thompson said. I think it's useful in evaluating how the items behave. For anyone that's interested, you can calculate it in R using the mirt package, e.g.:
itemplot(fit.ext.grm, item = 1, type = "score")
how can i draw the Item category response functions for a five‐category item in spss. please guide me
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software