If raw scores and rasch estimates correlate so well, why bother? Value adding in test development and scoring with the Rasch model

Xitao Fan (1998) wrote the oft-cited report of a large-scale empirical investigation which concluded, inter alia, that Rasch measurement results are not worth the extra effort necessary to produce them. In that case many might conclude that learning and implementing Rasch measurement techniques has no real benefit: raw scores and True Score Theory would seem to be both easier and adequate. Given that N is the sufficient statistic for estimating person and item parameters only to the extent to which the date fit the Rasch model, then data to model fit should confer some added value to test development and to data analysis. This presentation argues that Rasch quantity estimates of ability are interval level measures crucial for measuring difference and growth, while Rasch fit indicators provide the means to monitor measurement quality, and to diagnose the often overlooked inadequacies of item, person and test performances.