Skip to main content

Heteroskedasticity

Heteroskedasticity is a violation of the ordinary least-squares assumption that the variance of the regression error is constant across observations.

Also known asnon-constant variance · heteroscedasticity

ByHoang TruongUpdated

FrameworkOrdinary least squares (OLS)

See it move

Loading infographic...

A scatter chart plots regression residuals on the vertical axis against fitted values on the horizontal axis, with a horizontal reference line at zero. Instead of forming a uniform horizontal band, the scatter fans outward into a widening cone as fitted values increase — the visual signature of heteroskedasticity. This pattern shows that the error variance is not constant across observations, which makes OLS standard errors unreliable and invalidates conventional significance tests unless corrected.

Where it fits
SubjectData Analysis & StatisticsAdvancedTopicRegression Diagnostics & ProblemsAdvanced