Does regression analysis of lung function data obtained from occupational epidemiologic studies lead to misleading inferences regarding the true effect of smoking?