TPC Journal-Vol 11-Issue-3 - FULL ISSUE

274 The Professional Counselor | Volume 11, Issue 3 Table 1 Fit Indices and Tentative Thresholds for Evaluating Model Fit Absolute Fit Indices Index Abbreviation Strong Fit Acceptable Fit Poor Fit Chi-square χ2 or CMIN p-value > .05 or χ2 to df ≤ 1 χ2 to df ≤ 2 or 3 χ2 to df > 3 Standardized root mean square residual SRMR < .05 .06 to .08 > .08 Root mean square error of approximation RMSEA < .05, report confidence interval .06 to .08 (.081 to .10 can denote a somewhat acceptable fit) > .10 Goodness-of-fit index & Adjusted goodness-of-fit index GFI/AGFI ≥ .97 ≥ .95 (≥ .90 to .94 can denote a somewhat acceptable fit) < .90 Incremental Fit Indices Index Abbreviation Strong Fit Acceptable Fit Poor Fit Comparative fit index CFI ≥ .97 .95 to .90 < .90 Normed fit index NFI ≥ .97 .95 to .90 < .90 Incremental fit index IFI ≥ .97 .95 to .90* < .90 Tucker–Lewis index TLI ≥ .97 .95 to .90 < .90 Parsimonious Fit Indices Index Abbreviation Strong Fit Acceptable Fit Poor Fit Parsimony-adjusted goodness-of-fit index PGFI Parsimony-adjusted indices range from 0 to 1 and have utility for making comparisons between different models. Values closer to 1 indicate a stronger fit. Parsimony-adjusted normed fit index PNFI Note. The fit indices and benchmarks to estimate the degree of model fit in this table are offered as tentative guidelines for scores on attitudinal measures based on the synthesized recommendations of numerous psychometric researchers (see citations in the “Confirmatory Factor Analysis” section of this article). The list of fit indices in this table are not allinclusive (i.e., not all of them are typically reported). There is no universal approach for determining which fit indices to investigate nor are there any absolute thresholds for determining the degree of model fit. No single fix index is sufficient for determining model fit. Researchers are tasked with selecting and interpreting fit indices holistically (i.e., collectively), in ways that make both statistical and substantive sense based on their construct of measurement and goals of the study. *.90 to .94 can denote an acceptable model fit for incremental fix indices; however, the majority of values should be ≥ .95. Model Respecification The results of a CFA might reveal a poor or unacceptable model fit (see Table 1), indicating that the dimensionality of the hypothesized model that emerged from the EFA was not replicated or confirmed with a second sample (Mvududu & Sink, 2013). CFA is a rigorous model-fitting procedure and poor model fit in a CFA might indicate that the EFA-derived factor solution is insufficient for appraising the construct of measurement. CFA, however, is a more stringent test of structural validity than EFA,