Hu and Bentler (1998, 1999) extensively studied many indices and provide guidance for selecting and interpreting a manageable subset. They recommend reporting one residuals-based measure such as the standardized root mean square residual (SRMR; Bentler, 1995; Jöreskog and Sörbom, 1981), and one or more of the following: (a) the root mean square error of approximation (RMSEA; Browne and Cudeck, 1993; Steiger, 1990; Steiger and Lind, 1980), (b) the Tucker–Lewis (1973) incremental fit index (TLI; also known as the non-normed fit index due to Bentler and Bonett, 1980), (c) Bollen's (1988) non-normed index (Δ2), and (d) Bentler's (1990) comparative fit index (CFI). CFI is always between zero and 1. It avoids the underestimation of NFI and the overestimation of NNFI. In this category of fit indices, CFI is the most frequently reported one. https://www./topics/mathematics/fit-index |
|