Hierarchical regression
This example of hierarchical regression is from an Honours thesis – hence all the detail of assumptions
being met. In an undergraduate research report, it is probably acceptable to make the simple statement
that all assumptions were met.
3.2.2 Predicting Satisfaction from Avoidance, Anxiety, Commitment and Conflict
Prior to conducing a hierarchical multiple regression, the relevant assumptions of this statistical
analysis were tested. Firstly, a sample size of 96 was deemed adequate given five independent
variables to be included in the analysis (Tabachnick & Fidell, 2001). The assumption of singularity
was also met as the independent variables (Avoidance, Anxiety, Commitment and Conflict) were not a
combination of other independent variables. An examination of correlations (see Table 2) revealed that
no independent variables were highly correlated, with the exception of Conflict and Satisfaction.
However, as the collinearity statistics (i.e., Tolerance and VIF) were all within accepted limits, the
assumption of multicollinearity was deemed to have been met (Coakes, 2005; Hair et al., 1998).
Extreme univariate outliers identified in initial data screening were modified as above. An examination
of the Mahalanobis distance scores indicated no multivariate outliers. Residual and scatter plots
indicated the assumptions of normality, linearity and homoscedasticity were all satisified (Hair et al.,
1998; Pallant, 2001).
A four stage hierarchical multiple regression was conducted with Satisfaction as the dependent
variable. Social Desirability was entered at stage one of the regression to control for socially desirable
responding. The Attachment variables (Avoidance and Anxiety) were entered at stage two,
Commitment at stage three and Conflict at stage four. The Relationship variables were entered in this
order as it seemed chronologically plausible given attachment is relevant from infancy, whereas
commitment and conflict issues occur once a person is in a relationship. Intercorrelations between the
multiple regression variables were reported in Table 2 and the regression statistics are in Table 3.
