Hello! I have a question regarding simulation of many-facets Rasch data using the Facets simulation feature. I have a situation where I am attempting to add additional persons that are exhibiting certain rater effects (halo, range restriction, etc). I read the following link:
In example 2 it discusses anchoring the persons and using dummy scores for additional ratings and to model specific rater effects. However, I'm not sure how anchoring would create a halo effect pattern (4, 4, 4, 4 for example), since the fit statistics are not used as the anchor value.
Could you help to clarify this? Are fit statistics or response patterns used in anchoring?
Eli: to simulate non-Rasch behavior, we must simulate several datasets and combine them into one analysis. For instance, for a halo effect, you would Rasch-simulate one observation on the first item for each person for each rater, and then replicate this observation for the other items for that person and rater. Then combine this dataset with a dataset in which all the observations are Rasch-simulated.