Very nice video. I'm currently using the mirt package to run dif analysis on a bifactor model of dichotomous items. There's no good videos out there on how to do it and I think there are enough little specific considerations to make a whole series of videos on this topic. If you're looking for ideas, I would love to have some more resources on that topic so that I know I'm doing it correctly. Keep up the good work!
Thank you for the compliments. It is a great idea for follow up video. I will definitely try to work on that and will be posting it in the very near future.
Thank you very much for a great video! What is the link between the MH chi square statistic and the effect size (A-C)? In my dataset the item with the largest MH-chi square statistic (629,49) has an effect size in the category "B", while another item has a MH-chi square statistic of 56.14 and is in the effect size category "C". I hope you can shed light on this mystery for me :)
Greetings, thank you for your valuable contribution. I have tried to run the data from an exam. According to your script, but I can't move forward because that error appears. what can i do? > test2
It could be that there is an NA value (i.e., missing value) where an actual value is needed. I reviewed the notes in the function and it indicates that -- "Missing values are allowed for item responses (not for group membership) but must be coded as values. They are discarded for item parameter estimation." Try reviewing and seeing if you have NA values in group membership and if so taking them out.
Thanks for a nice video. May I have a question? I am currently working with a dataset with 6 different groups. I am wondering how to analyze DIF among 6 different groups because normally people only compare 2 groups at one time. I appreciate any suggestion from you about how to do DIF in my case using R. Thank you in advance.
That is a really good question. If you want to keep it more simple one option is that you can consider categorizing all your groups into two groups. For example if you have 6 schools you categorize them into public and private schools. But this may not be feasible or it may move away for your intended purpose. Now the package does allow and will recognize multiple groups. However the interpretations may differ now. Magis et al (2010) pg 848 under the section 'Number of focal groups' provides a really good description. Here is the article: link.springer.com/content/pdf/10.3758/BRM.42.3.847
Hello and thanks for the compliment. Quick question, did you also install the mirt package you can install it with install.packages("mirt")? If not, try installing it first before you run the line library(mirt).
Here is the R code:
# where to download IQ1 data
# openpsychometrics.org/_rawdata/
# load data downloaded into R
IQ1
Very nice video. I'm currently using the mirt package to run dif analysis on a bifactor model of dichotomous items. There's no good videos out there on how to do it and I think there are enough little specific considerations to make a whole series of videos on this topic. If you're looking for ideas, I would love to have some more resources on that topic so that I know I'm doing it correctly. Keep up the good work!
I agree, mirt content would be great.
Thank you for the compliments. It is a great idea for follow up video. I will definitely try to work on that and will be posting it in the very near future.
very nice video man, it helped me a lot!!, thanks!
I am happy you found it helpful.
It was perfect, Thank you, learned a lot :)
Thanks so much for that complimented. It is very much appreciated.
Thank you very much for a great video! What is the link between the MH chi square statistic and the effect size (A-C)? In my dataset the item with the largest MH-chi square statistic (629,49) has an effect size in the category "B", while another item has a MH-chi square statistic of 56.14 and is in the effect size category "C". I hope you can shed light on this mystery for me :)
Greetings, thank you for your valuable contribution. I have tried to run the data from an exam. According to your script, but I can't move forward because that error appears. what can i do?
> test2
It could be that there is an NA value (i.e., missing value) where an actual value is needed. I reviewed the notes in the function and it indicates that -- "Missing values are allowed for item responses (not for group membership) but must be coded as values. They are discarded for item parameter estimation." Try reviewing and seeing if you have NA values in group membership and if so taking them out.
Thanks for a nice video. May I have a question? I am currently working with a dataset with 6 different groups. I am wondering how to analyze DIF among 6 different groups because normally people only compare 2 groups at one time. I appreciate any suggestion from you about how to do DIF in my case using R. Thank you in advance.
That is a really good question. If you want to keep it more simple one option is that you can consider categorizing all your groups into two groups. For example if you have 6 schools you categorize them into public and private schools. But this may not be feasible or it may move away for your intended purpose. Now the package does allow and will recognize multiple groups. However the interpretations may differ now. Magis et al (2010) pg 848 under the section 'Number of focal groups' provides a really good description. Here is the article:
link.springer.com/content/pdf/10.3758/BRM.42.3.847
Great Video! Any Idea why I get:
ERROR: dependency 'mirt' is not available for package 'difR'
when trying to install difR package?
Thanks.
Hello and thanks for the compliment. Quick question, did you also install the mirt package you can install it with install.packages("mirt")? If not, try installing it first before you run the line library(mirt).