The constant variance part is important though. If you apply linear regression to a severely heteroscedastic dataset, the inferences you obtain from the model w

The constant variance part is important though. If you apply linear regression to a severely heteroscedastic dataset, the inferences you obtain from the model will be incorrect.