[Rasch] a query on comparability across time
liasonas at cytanet.com.cy
Mon Mar 19 01:08:34 AEDT 2018
It seems that I need some of your lights, again!
One of you may have faced this problem in the past, and you may already know the answer to my question.
My situation: I have a few thousand candidates who have been examined on a number of tests. They were all examined on the Language test. They all took one of three 'Skills' tests. They all took one of several other tests (so three tests per person overall).
Assume that I have a dataset where each person is a row and each test is a column. Each person has score on three tests (columns). The maximum score for any of the columns is 100. It is as if I have a long test with many items, each item scored on a rating scale from 1-100. The data is connected, through the Language and Skills tests, so no disjointed data problem.
So far, so good, I can run a Rating Scale Rasch model. However, how can I compare this group with the next group of candidates, two years later? The tests will be different (no possibility to have similar questions) and the group of candidates will be different. I understand that this is partly a psychometric and partly a political problem (e.g. local politics do not allow any formal equating activity through secret common items etc).
Thank you for your time, I hope one of you has solved a problem like this in the past.
Thank you again
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch