When the spread of a regression's errors changes across the range of the inputs instead of staying constant, which is called homoscedasticity. The coefficient estimates stay unbiased, but the usual standard errors, and so the p-values and confidence intervals, become unreliable.
Card 1 of 244. Answer: When the spread of a regression's errors changes across the range of the inputs instead of staying constant, which is called homoscedasticity. The coefficient estimates stay unbiased, but the usual standard errors, and so the p-values and confidence intervals, become unreliable.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
Ordinary least squares assumes every error has the same variance. Heteroscedasticity breaks that assumption: the errors are noisier for some values of the inputs than for others. The coefficients are still unbiased, but OLS is no longer the most efficient estimator, and the textbook formula for standard errors is wrong, often understating the uncertainty.
Example: regress household spending on income. Low earners spend most of what they earn, so their residuals are small; high earners range from frugal to lavish, so the residuals fan out as income grows. A plot of residuals against fitted values shows a funnel instead of an even band.
To detect it, look at that residual plot or run the Breusch-Pagan or White test. To handle it, use heteroscedasticity-robust standard errors (often called White or HC standard errors), transform the target, for example by taking logs, or fit weighted least squares with weights inversely proportional to the variance. The follow-up question is usually whether it biases the coefficients: it doesn't.
