Heteroscedasticity is the condition that a collection of random variables does not have a common variance. In a regression model, the errors are heteroscedastic when their conditional variances vary with the values of the independent variables, as in
 |
where 
denotes the independent-variable data and the 
are not all equal. Heteroscedasticity affects formulas
that assume a common error variance,
including the usual standard error formulas for
ordinary least squares fitting.
