[Rasch] Rasch vc CFA

Stone, Gregory Gregory.Stone at UToledo.Edu
Sat Sep 29 12:51:23 EST 2012

My broader concern in this discussion is why?  Why are we comparing Rasch to CFA or FA for that matter?  One is a measurement model, the other is a descriptive correlational investigative tool.  Thurstone's factor analysis presupposes the existence of a measured variable to begin the evaluation process, not as an outcome of the evaluation.  Guildford's classic 1941 article in Psychometrika illustrates the mistakes associated with assuming measures are products of factor analyses, and by extension into the CFA variable molding process.

As a tool for revealing potential aspects within a measure, factor analytic techniques have some merit, but they was never designed to measure.  Screwdrivers make poor hammers, and we can't measure with factor analysis.  Thurstone's partner on operationalizing FA, Truman Kelley, wrote what has become my standard explanation for FA/CFA: "There is no search for timeless, spaceless, populationless, truth in factor analysis; rather, it represents a simple, straightforward problem of description in several dimensions of a definite group functioning in definite manners, and he who assumes to read more remote verities into the factorial outcome is certainly doomed to disappointment."


Gregory Stone
Research and Measurement
The University of Toledo
Toledo, OH 43606

More information about the Rasch mailing list