[Rasch] effects of misfitting persons

Mike Linacre rmt at rasch.org
Wed Feb 1 16:29:53 EST 2012


Thank you for this question, Jason:

"The general question is: how does the inclusion of misfitting students 
affect the quality of measurement for other facets e.g. examiners?"

The effect of huge misfit is usually to decrease the accuracy and logit 
range of all the measures.

But why not answer this question for yourself ?

1) Perform an analysis of all your data.
2) Clean up your data and reanalyze.
3) Cross-plot the measures for the examiners from 1) and 2)

We expect that the measures from 1) and 2) will be highly correlated. The 
examiner measures from 2) will have a wider range than the measures from 1).

Please experiment with the Facets example file, Dives.txt. It includes one 
highly misfitting diver (diver 7 on dive 8 who did the wrong dive). You can 
see the effect on the measures of analyzing the data with and without him.

OK?

Mike L.




More information about the Rasch mailing list