A New Test for Heteroskedasticity

摘要: Abstract The quite general test for heteroskedasticity presented here regresses the absolute values of residuals obtained by ordinary least-squares on some variable(s). Denoting O.L.S. , one obtains, instance, a regression like where z is variable, and b coefficients new regression. We call acceptance non-zero value both case “mixed heteroskedasticity”, which we deem frequent in practice though neglected handbooks. paper also summarizes another due to S. M. Goldfeld R. E. Quandt examines powers two using Monte-Carlo simulations: seems compare favourably, except perhaps large samples.