TPC Journal-V6, Issue 4- FULL ISSUE

The Professional Counselor /Volume 6, Issue 4 350 Table 1 Correlations among measures of direct student services, perceived stress, and burnout M SD 1 2 3 4 5 6 7 8 9 1. Counseling 3.02 .60 - 2. Curriculum 2.77 1.16 .44 - 3. Percent of Time 59 78 .36 .27 - 4. Perceived Stress 1.56 .63 -.15 -.11 -.14 - 5. Exhaustion 3.04 .86 -.15 -.11 -.11 .61 - 6. Incompetence 2.29 .68 -.31 -.14 -.18 .49 .44 - 7. NEW 2.56 .87 -.23 -.19 -.22 .46 .53 .39 - 8. DC 1.39 .50 -.20 -.17 -.14 .32 .28 .45 .64 - 9. DPL 2.39 .80 -.19 -.12 -.16 .58 .66 .41 .47 .30 - Note. N = 926. All correlations ( r ) were statistically significant ( p < .001). Counseling = frequency of direct counseling services, curriculum = frequency of direct curriculum services, percent of time = percent of time in direct services to students, NEW = negative work environment, DC = devaluing client, DPL = deterioration in personal life. Multiple fit indices were examined to determine the goodness of fit for the measurement model and structural model (Hu & Bentler, 1999; Kline, 2011; Weston & Gore, 2006). The fit indices that were used include: (a) chi-square, (b) comparative fit index (CFI), (c) goodness of fit (GFI), (d) standardized root mean square residual (SRMSR), and (e) root mean square error of approximation (RMSEA). Furthermore, we consulted the normed fit index (NFI) and Tucker-Lewis index (TLI) because they are more robust to non-normal data as compared to other indices (Lei & Lomax, 2005). For a detailed description of these fit indices, readers can review the works of Hu and Bentler (1999), Kline (2011), and Weston and Gore (2006). We used these fit indices to establish a diverse view of model fit. Measurement model . First, we employed a CFA model to examine the latent variable representing burnout (Lee et al., 2007). The research team totaled each subscale on the CBIs to develop a composite score for each domain. The initial measurement model for burnout produced acceptable standardized factor loadings ranging from .41 (devaluing client) to .57 (incompetence), .62 (negative work environment), .77 (deterioration in personal life), and .82 (exhaustion). Furthermore, all fit indices for the measurement model indicated an adequate fitting model except chi-square, RMSEA, and TLI: χ2 ( df = 5, N = 926) = 107.07, p < .001; GFI = .96; CFI = .92; RMSEA = .15; SRMR = .06; NFI = .92; TLI = .85. Therefore, we consulted the modification indices and standardized residual covariance matrix and tested a new CFA based upon these consultations. The modifications indices indicated the need to correlate the error terms for incompetence and devaluing client. The resulting model produced a model in which all fit indices indicated an adequate fitting model: χ 2 ( df = 4, N = 926) = 12.03, p = .02; GFI = .99; CFI = .99; RMSEA = .05; SRMR = .02; NFI = .99; TLI = .99. Further inspection of the standardized factor loadings for the model indicated they were all acceptable except for the factor loading for devaluing client, which dropped to .36 (below .40; Stevens, 1992). While these modifications improved the overall fit of the CFA, the correlation of incompetence and devaluing client has no theoretical justification (Byrne, 2010). In addition, the correlation of the error terms for incompetence and devaluing client produced a standardized factor loading below the noted standard of .40 (Kline, 2011; Stevens, 1992). Subsequently, we removed the