What is Heteroscedasticity?
Heteroscedasticity refers to data for which the variance of the dependent variable is unequal across the range of independent variables. Heteroscedasticity is the opposite of homoscedasticity. The heteroscedasticity of data is important in the context of regression analysis. A regression model assumes a consistent variance, or homoscedasticity, across the data.
Heteroscedasticity in the data results in regression providing accurate outputs on one end of the data range but highly inaccurate outputs on the other end of the data. An easy way to visualize these concepts is to create a scatter plot of the data. A heteroscedastic data set will exhibit a conical shape across the range of independent variables. The wider the cone, the more heteroscedastic the data is and the less friendly for regression analysis. It is important to understand that a regression analysis on the data set is still possible but the results will prove unreliable outside of a specific range.