[Rasch] computing outfit mean-squares for each rating category in eRm (PCM)

Alexandra Dima alexadima at gmail.com
Tue Aug 9 19:40:05 AEST 2016


Dear Trevor, 


Thanks for your question, I am following the example from this eRm vignette:


On page 16, they show how to perform a likelihood ratio test between RSM and PCM. As in the example there, the test on my dataset is significant. 

I understand this means that the distances between category thresholds are not equal – if I free up those parameters the data fits the model better. I know that’s a lot of free parameters (31 vs 10 in my case). 

Should I consider other criteria when choosing between RSM and PCM?  Can you recommend a text where this is explained in more detail?


Regarding the outfit MSQ for rating categories, the authors of the paper I try to replicate follow this source:


Both RSM and PCM are mentioned here, no advice on model choice (paper has a different focus).

The paper I replicate does not mention which model they used. 



Many thanks, 




From: Rasch <rasch-bounces at acer.edu.au> on behalf of "Bond, Trevor" <trevor.bond at jcu.edu.au>
Reply-To: <rasch at acer.edu.au>
Date: Tuesday 9 August 2016 at 10:28
To: "rasch at acer.edu.au" <rasch at acer.edu.au>
Subject: Re: [Rasch] computing outfit mean-squares for each rating category in eRm (PCM)


Dear Alex

In what sense do you judge the PCM to have better fit? The RSM is much more parsimonious. Remember Occam's Razor.



Sent from 007's new iPhone 6s ��

On 8 Aug 2016, at 5:16 PM, Alexandra Dima <alexadima at gmail.com> wrote:

Dear Rasch colleagues, 


I am trying to replicate a Rasch analysis performed with WINSTEPS. 

They examine the item rating structure of an 8-item questionnaire (5-point Likert response format). 

One criterion they consider is that the outfit mean-squares are <2.0 for each rating category. 


I am using the eRm package in R – the RSM and PCM functions. 

PCM fits better, so I use it for examining item properties etc.

I can find item outfit and infit, but not outfit mean-squares for each rating category.

There must be a way to get it (I see the infit t statistic for each category in the Item Map via the plotPWmap function).


Any suggestions on how I can compute these?


Many thanks, 


Rasch mailing list
email: Rasch at acer.edu.au
web: https://mailinglist.acer.edu.au/mailman/options/rasch/trevor.bond%40jcu.edu.au

________________________________________ Rasch mailing list email: Rasch at acer.edu.au web: https://mailinglist.acer.edu.au/mailman/options/rasch/alexadima%40gmail.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20160809/1ddb48c9/attachment.html>

More information about the Rasch mailing list