[Rasch] Modeling sequenced data

Mike Linacre (RMT) rmt at rasch.org
Sat Dec 8 21:37:56 EST 2007


Thank you for asking for our input. Your sequence task appears to be 
equivalent to rank-ordering the statements.

A straight forward approach is to conceptualize the statements as "items", 
each complete ordering as a "person" row, and the data as Andrich rating 
scale data on a rating scale from 1 to 10.

It can be shown that this approach introduces some bias into the estimates 
relative to a strict Rasch rank-order model, but the difference is not 
meaningful for practical purposes. See JAM 7:1 p. 129-136.

Mike Linacre

At 12/7/2007, you wrote:
>My colleagues and I have been collecting data with an instrument called 
>the "sort task" for several years. The task involves sorting a set of ten 
>statements into the order in which it would have been possible for one 
>individual to have made them over the course of 40 years. (They end up 
>being ordered from 01 to 10.)
>We now have hundreds of cases, and would like to model the results. We 
>already know that the modal order for each group we have tested is 
>virtually always accurate, and that higher level statements are more 
>difficult to order than lower level statements there must be a way to use 
>the Rasch model to observe the difference in difficulty between sorting 
>lower level statements and higher level statements. We also have data on 
>age, education, and the techniques people use to do the sort and will be 
>examining the relation between these variables and sort accuracy.
>I've never tried to model data of this type. Can anyone help?

More information about the Rasch mailing list