Pitfalls in using the Goldfeld-Quandt test against heteroscedasticity: a comment on