Table 2.

Regression coefficients and odds ratios from a model predicting intention to leave the field of early childhood education^a

Measure	B (SE)	Adjusted Odds Ratio (95% Confidence Interval)
Work as a calling scoreb	– 0.09 (0.03)*	0.91 (0.85, 0.97)
Race/ethnicity
White, non-Hispanic	Reference	Reference
Black, non-Hispanic	– 0.15 (0.47)	0.86 (0.34, 2.15)
Other	– 1.56 (1.07)	0.21 (0.03, 1.70)
Serious ongoing stress at work
No	Reference	Reference
Yes	1.33 (0.78)	3.78 (0.82, 17.32)
Constant	1.10 (1.49)	3.02 (0.16, 56.42)

^*p < .01

^a N = 191. There was listwise deletion of 3 participants who were missing data on race/ethnicity. For overall model evaluation, the Wald test was used and indicated that the logistic regression model with all three predictors was an improvement over the intercept-only model (χ² (4) = 15.56, p = .004). For goodness-of-fit, the Hosmer-Lemeshow test was used and was insignificant, indicating that the model fit the data well (χ² (8) = 4.52, p = .807)

^b In the logistic regression model, with calling score as the predictor variable and intention to leave the early childhood education field as the dependent variable (and with no covariates), for every one point increase in the calling score, the odds of intention to stay decreased from 1.00 to 0.90 (95% CI: 0.85, 0.96). In the logistic regression model, after adjusting for race/ethnicity and stress at work, the addition of the calling score to the model significantly improved model fit, as assessed by the Wald test (χ² (1) = 8.20, p = .004)