Multiple Regression with WINSTEPS: A Rasch Solution to Regression Confusion

Abstract

The purpose of Rasch measurement is to build and verify a useful “yardstick” – a stable, portable, reproducible instrument for making linear measures. What makes a yardstick useful is the calibration of its reference points, which mark out a visible linear metric that maintains its spacing as long as it is used in a sensible way. Just like the yardstick in your closet, a useful “yardstick” does not change the distances between its calibration marks from object to object, place to place, or time to time, as long as you apply it as sensibly as you apply the yardstick in your closet. What follows explains how to use the Rasch measurement program, WINSTEPS (winsteps.com,
Linacre, 2000) to solve multiple regression problems in a new way that avoids the sample covariance dependence and missing data problems which interfere with inferential stability.