However, the base model (figure 1) was partially supported with standardized coefficients. The Chi-square X2 (df = 04, N = 268) = 82.5, p<.001 reflected that the model was not a good fit to the data. The fit indices used in the study consist of commonly reported Comparative Fit Index (CFI) and Root Mean Square Error of Approximation (RMSEA). The CFI index compares the base model with the independence model: the values greater than .95 indicates a good-fitting model (Hu & Bentler, 1999). The RMSEA assesses the comparability of the base model with an ideal (saturated) model. The more it approximates zero, the better the model would be. However, the values below .08 suggest a good fit. The hypothesized model (figure 1) displayed a poor CFI of .737 and RMSEA of .271. Another indication of the goodness of fit lies in the standardized residuals, which expected to be less than .080 in order to have a good-fit model. Present base model (figure 1) yielded an unacceptable value of standardized residuals (.113) which was beyond the acceptable limit. Likewise, remaining fit indices such as goodness of fit index (GFI, measures the effectiveness of the model in approximating the observed covariance matrix, cut off value >.95), created by Jöreskog and Sörbom (1998), reflected a satisfactory GFI value (.907). However, the adjusted goodness-of-fit index (AGFI, acceptable value >.90), Tucker and Lewis’s non-normed fit index (TLI, acceptable value >.90) were also found to be unacceptable: .511 and .150 respectively, for this model. The ratio of Maximum-Likelihood Chi-Square to the degrees of freedom (X2/df, acceptable value < 5) was also found to be unacceptable 20.62.
            In order to develop a better-fitting model, some Post Hoc Model modifications were conducted successively. At first, Guilt was allowed to predict Attribution and further, the error variances of Realization & Reparation and Guilt, as well as Positive and Negative Self-compassion, was made correlated. However, to obtain a perfect fit model one more modification was done by removing the insignificant predictive pathway from the model. The pathway leading from Self-Esteem to Guilt was removed subsequently. Only a single modification was adapted at a time. These modifications produced a perfect fit model: X2/df (df =2) =1.566; CFI = .996; RMSEA = .04; SRMR=.017; GFI=.996, AGFI= .959 (Figure 2).
[FIGURE 2 MAY BE PLACED HERE]
The standardized direct path coefficients of the tested model (figure 2) are presented in table 2.
Table 2. Direct effects of self-esteem and self-compassion on the dimensions of self-forgiveness
Estimate
Negative self-compassion <--- Self esteem .27**
Positive self-compassion <--- Self esteem .52**
Guilt <--- Negative self-compassion .21***
Guilt <--- Positive self-compassion -.12*
Attribution <--- Self esteem -.21*
Attribution <--- Negative self-compassion -.19***
Realization & reparation <--- Positive self-compassion .34*
Attribution <--- Positive self-compassion .20*
Realization & reparation <--- Self esteem .16*
Realization & reparation <--- Negative self-compassion -.18***
Attribution <--- Guilt -.183
**p<.01, *p<.05