I have encountered several instruments in psychology and related fields where we have multiple versions of the same instrument, specifically a short form and a long form. For instance, the patient health questionnaire has a 8 item version and a 9 item version. As another example, the perceived stress scale has a 4 item version, 10 item version, and 14 item version. I am looking at a scale with a 50 item long version and a 10 item short version. I am wanting to develop some formula or conversion table to allow me to go from one form of a test to anther.
I was thinking of approaching this problem using test equating, specifically Common-Item Non-Equivalent Group Designs. But there seems to be a critical nuance with this particular situation. The common items that we see on the long form of the test are all the items on the short form of the test.
Is equating the best way to approach this, or would reviewers want to see a regression model/predictive formula? Does anyone know of a research study that attempted to do something similar to this?
