[Rasch] Equating tests in RUMM2020 using anchors

Stephanou, Andrew Andrew.Stephanou at acer.edu.au
Wed Jul 25 08:38:03 EST 2012



Sent to the list on behalf of Matt Homer



-----Original Message-----
From: Matthew Homer [mailto:M.S.Homer at education.leeds.ac.uk]
Sent: Tuesday, 24 July 2012 7:17 PM
To: rasch
Subject: Equating tests in RUMM2020 using anchors

Folks,

I am trying to link two tests using anchors - I have approximately 310 items in the first test, and 350 in the second, with 40 common anchors. The number of test-takers is around 250 in each case.

I have got the analysis working to some extent in RUMM2020 when I run the whole lot together (all items, all respondents) - for example, the item stats and person stats look fine. However, I am getting errors when I try various other analyses (e.g. equating tests: "RUMM2020 is unable to complete plot display"; residual correlations: "sMsgStem as the Error: has occurred"!).

I have two questions:

1. Am I witnessing some sort of limitation with RUMM2020 - I just have too many items/test-takers? Or are there things I could do to try and get it working?

2. What are the pros and cons of my approach (all items, all respondents) with the more 'standard' approach of using the first test to calibrate the items and then importing these into the second analysis as fixed estimates? Any references on this latter issue would be most gratefully received.

I hope my question is clear.

Many thanks,


Matt Homer
Research Fellow
Schools of Education and Medicine
University of Leeds

http://www.education.leeds.ac.uk/modx/people/staff/academic/homer

Usual office hours:
School of Education (0113 343 4654): Monday, Wednesday, Thursday
School of Medicine (0113 343 4304): Tuesday, Friday

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailinglist.acer.edu.au/pipermail/rasch/attachments/20120724/08a4b7dd/attachment.html 


More information about the Rasch mailing list