ABSTRACT
Some of Rod McDonald’s most widely cited articles (McDonald, 1989; McDonald &
Marsh, 1990; also see Marsh, Balla, & McDonald, 1988) deal specifically with assessing
goodness of fit (GOF) and related earlier work on unidimensionality and the nonrelation
between coefficient alpha and dimensionality (e.g., McDonald, 1981)—despite his
expressed ambivalence about the use of GOF indices. Indeed, their continuing popularity
may have more to do with the Zeitgeist that existed in structural equation modeling
research-and continues to exist today-than in his enthusiasm for their routine
application. Rod was dragged, somewhat reluctantly, into this area of inquiry by one of
his former PhD students (John Balla), who had been enticed by another Australian
colleague (H.W.Marsh) to this topic. Applying diligence, imagination, lateral thinking,
and a bit of naivete, Marsh and Balla were able to find or devise a plethora of alternative
goodness-of-fit indices-32 of them-and a preliminary classification schema (absolute,
Type 1 incremental, and Type 2 incremental) for generating and classifying these indices.
Because many of these indices were not readily amenable to mathematical scrutiny, they
constructed a simulation study (based on real data, and artificial data fit to “true” and