A Viable Measurement Approach for Model Comparison

Abstract

The debate about the use of the theory of change scores has spewed in the academic literature. As a result, the use of a statistical method is developed based on these theories are indirectly questionable. The aim of the current study was to explore the simple gain score and the residual gain score methods to unearth the reasons for the Lord’s paradox is the backbone of the current debate. Using a Monte Carlo Simulation approach, the simple gain score and the residual gain score approaches produced inconsistent results mainly due to the pre-test differences and covariate independence. A mixture of the group-centred transformation and the propensity score matching approach is proposed as a way of improving the accuracy and interpretation of the results. However, it is not guaranteed that the results will be less biased in panel data where self-selection is the default selection criteria.