When exposure is subject to nondifferential misclassification, are validation data helpful in testing for an exposure–disease association?