[Rasch] Removal of Participants

Mike Linacre rmt at rasch.org
Thu Jan 12 14:33:46 EST 2012


Thank you for question, J.

The general rule is: "When in doubt, analyze the data both ways. Then base 
your decisions on the analysis that makes the most sense."

See, for instance, http://www.rasch.org/rmt/rmt234g.htm

Mike Linacre

At 1/11/2012, you wrote:

>Dear all,
>
>I have a large sample size of n=3798 persons to Rasch analyze a survey 
>instrument of 14 items (with 4-point ordinal response options.).  Using 
>Winsteps, I noticed that the person measures ranged from -12.16 to 7.16 
>logits.  I noticed further that around 500 persons with logits/measures 
>of  greater than 7.0 have outfit and infit mean squares of zero..Many 
>persons (around 800) have outfit and infit mean squares outside the 0.5 
>to1.5 range.  Is it a good practice to remove or exclude the persons from 
>the analysis those with zero outfit and infit mean squares?
>
>Your inputs are highly appreciated.
>
>Thank you.
>J




More information about the Rasch mailing list