A note on Bayesian score tests for heteroscedasticity