Correcting measurement errors with Bayesian networks when the variables are continuous and discrete