Hierarchical Regression Analyses

Podcast -- [Hierarchical Multiple Regression -- SPSS~12 mins] 

The following illustration is from the work of Dr. Wanda Briggs.

Wanda wanted to know if multicultural education and training contributes to counselors' ethnocultural empathy. Previous research and theory suggested that counselors' demographic characteristics (age, race, and age) as well as their diverse experiences should influence counselors' ethnocultural empathy. She wants to run a hierarchical regression to examine the amount of variability in ethnocultural empathy that education and training accounts after controlling for demographic characteristics and diverse experiences. A predetermined level of entry of the independent variables was performed to allow the researcher to examine the unique contribution above, and beyond each group of independent variables.  Two categorical variables were coded to represent a binary code.  Binary codes established were gender (coded 1 if female and 0 if male), and race/ethnicity (coded to represent 1 if White and 0 otherwise).  In first step of the hierarchical regression procedure, the demographic predictor variables, age, gender, and race/ethnicity were block entered providing the variance accounted for in this group of independent variables. In the second step diversity experience was entered into the step 1 model.  Next, the researcher repeats the entry of the first and second step and block enters the education and training predictor variables, including reported academic credit hours, multicultural academic credit hours, and MCI subscale scores: skill, awareness, relations, and knowledge providing for the unique contribution of education and training to the dependent measures. 

[Download Data]



After you finish entering the final block of variables, go to Statistics.



Selected SPSS Output



Prior to the hierarchical regression analyses, the independent variables were examined for collinearity.  Results of the variance inflation factor (all less than 2.0) suggest that the estimated bs are well established in the following regression models.

        The first research question addressed was:  Does multicultural training and education account for a significant amount of variability in enthnocultural empathy over and above that accounted for in demographic characteristics (age, gender, and race/ethnicity) and diverse experience?  The results of the hierarchical regression predicting enthnocultural empathy from demographic characteristics, diverse experience, and multicultural education and training are reported in Table 4. The results of step one indicated that the variance accounted for (R2) with the first three predictors (age, gender, and race/ethnicity) equaled .09 (adjusted R2=.08), which was significantly different from zero (F(3, 186)=6.45, p<.001).  Next, diversity experience scores were entered into the regression equation.  The change in variance accounted for (DR2) was equal to .14, which was statistically significant increase in variance accounted for over the step one model (DF(1,185)=14.32, p<.001).   In step three, education and training variables (academic hours, multicultural academic hours, and MCI subscale scores) were entered into the regression equation. The change in variance accounted for (DR2) was equal to .27, which was a statistically significant increase in variance accounted above the variability contributed by the previous predictor variables entered in step two (DF(6, 179)=15.93, p<.001).   Two of the demographic characteristics, age and race, and diverse experience were statistically significant. All the multicultural subscales were statistically significant, but the hours and multicultural hours were not statistically significant.  

Table 4

Hierarchical Regression Analyses Evaluating Predictors of Enthnocultural Empathy


















3, 186













































1, 185













Mulicultural Experience







6, 179




































































Note. Betas reported are those from the step at which the variable was entered into the equation.

*p< .05. ***p < .001.