ABSTRACT
When evaluating the goodness of fit of structural equations models (SEMs), researchers rely in part on subjective indices of fit as well as a variety of other characteristics (for more extensive overviews of assessing goodness of fit, see Bentler, 1990; Bentler & Bonett, 1980; Bollen, 1989b; Browne & Cudeck, 1989; Cudeck & Browne, 1983; Gerbing & Anderson, 1993; Marsh, Baila, & McDonald, 1988; McDonald & Marsh, 1990; Tanaka, 1993). There are, however, a plethora of different indices with no consensus as to which are the best. Adding to this confusion, major statistical packages (e.g., LISREL8, EQS, CALIS: SAS) tend to be overinclusive in their default presentation of indices, automatically including some that are known to have undesirable properties. Because there is no “best” index, researchers are advised to use a variety of qualitatively different indices from different families of measures (Bollen & Long, 1993; Tanaka, 1993). Whereas there is no broadly agreed upon typology of indices, the family of incremental fit indices is one of the most popular.
