Violations in homoscedasticity may be caused by non-normality in the underlying distribution of either \(X\) or \(Y\) scores or by the indirect effect of a third variable, and typically result in confidence intervals that are either too wide or too narrow.