Heteroscadasticity

. . has never been a reason to throw out an otherwise good model.