Raw Scores are NOT Linear Measures: Rasch vs. Classical Test Theory CTT Comparison

Raw Scores Are Not Linear Measures: : Rasch vs. Classical Test Theory CTT Comparison
Standing	Raw Scores	Linear Measures
Psychometrics	Classical Test Theory CTT	Rasch Model RM
Data	must be "complete"	robust against missing data and non-administered items
Meaning	ordinal ranking on hoped-for latent variable	linear positioning on latent variable explicitly-defined by item content
Status	mistaken for truth	known as estimates
Additivity	non-linear, bent	linear, straight
Continuity	discrete, lumpy	continuous, smooth
Precision	unknown, except on average	quantified by standard errors
Accuracy	unknown	quantified by local fit statistics
Linear Analysis	unsuited to usual statistics	ideal for usual statistics
Validity	sample-dependent test "reliability"	item-dependent construct validity
Diagnosis	sample-dependent item point-biserial	individual item and person fit statistics
Conceptualization	concrete	abstract
Occurrence	accidental	deliberate
Construction	irreproducible	reproducible
Scope	local	general
Relationships	incomparable	comparable
Context	test-bound	test-free
Integration	awkward to equate	simple to equate
Benjamin D. Wright

Raw Scores Are Not Linear Measures: : Rasch vs. Classical Test Theory CTT Comparison. B.D. Wright … Rasch Measurement Transactions, 1992, 6:1, 208