*3.4. Confirmatory Factor Analysis of CBB Model and CBB-R Model*

Table 4 analyzes the fit of the various models of the questionnaire by the original CBB model, the unidimensional CBB model, the proposed four-factor CBB model, and the revision of that model first proposed (CBB-R). The original model and the unidimensional model show values that are not very adequate. The proposed four-factor CBB model, which corresponds to what was found in the exploratory analysis, is better, but while it showed good fit indices, the advisable respecifications were made considering theoretical and statistical criteria (modification indices, errors of estimation, standardized errors of measurement), which led to elimination of items 2, 16, 3, 13, 17, and 11. The revised model showed much better fit with the calibration sample. The difference between the AIC default model value (248.497) and the AIC saturated model value (240.000) is also very small, showing that this is probably the best of the models according to the Akaike model selection criteria.

Fit indices for the proposed CBB-R model with the validation sample (*n* = 635) are shown in Figure 2. Confirmatory factor analysis for the proposed model was done, taking the following fit indices as measures: *χ*2/*gl =* 2.241, CFI = 0.961, TLI = 0.951, RMSEA = 0.044 (0.036–0.053).


**Table 4.** Fit indices for the models proposed (calibration sample; *n* = 605).

Note: CFI, comparative fit index; TLI, Tucker–Lewis index; RMSEA, root mean square error of approximation; CI, confidence interval; *df*, degrees of freedom; Est., estimation; Bel., below; Abv., above. CBB: CBB-R (Revised CBB Model).

The reliability of the model was analyzed using Cronbach's alpha, where α = 0.89 for the total sample; for factor 1 (job dissatisfaction), comprising four items, α = 0.697; for factor 2 (social climate), made up of three items, α = 0.666; for factor 3 (personal impact), made up of four items, α = 0.808; and for factor 4 (motivational exhaustion), comprising four items, α = 0.529. Furthermore, the data found by split halves also showed both equal-length (Spearman–Brown coefficient = 0.818) and unequal-length (Spearman–Brown coefficient = 0.819) consistency of the scales.

**Figure 2.** Proposed CBB-R model (validation sample *n* = 635). Note: F1: job dissatisfaction; F2: social climate; F3: personal impact; F4: motivational exhaustion.

Table 5 shows the values for all six models. It can be seen how ΔCFI is over 0.01 for models 3 and 4, accepting the configural and metric invariance. Specifically, ΔCFI between model 1 (configural and metric base model) and models 3 and 4 is 0.024, so scalar and strict invariance cannot be accepted. In the analysis of variance by gender, in all cases ΔCFI is under 0.01, so the configural, metric, scalar, and strict invariances are accepted.


**Table 5.** Multigroup analysis of variance by type of contract (permanent/temporary) and gender (male/female).

Nota FS = factor saturation; Int = intercepts; Err = error.
