Homoscedasticity

Synonyms

Assumption of normality; Homogeneity of variance

Antonyms

Heteroscedasticity

Definition

In statistics, homoscedasticity occurs when the variance in scores on one variable is somewhat similar at all the values of the other variable.

Description

To illustrate homoscedasticity, assume a group of researchers are collecting continuous data (i.e., correlation design), whereby they are looking at 10 year-olds’ weight and their stress level (as measured on a scale of 1–10). The data is said to be homoscedastic if the variance in the stress scores is somewhat the same across the children who weigh 70, 85 and 90 pounds.

In regression analysis, the assumption of homoscedasticity occurs when at each level of the predictor variable, the residuals have similar variances [1, 2].

One way to test this assumption in regression analysis is through requesting a residuals scatterplot. The predicted scores lie on the first axis of this scatterplot, while the residuals (prediction errors) lie on the...