ABSTRACT
The selection problem Identification Sample selection bias arises when there are missing data for the dependent variable of interest. Recall the discussion of item non-response in the introduction. For example, in the HALS data set, measures of physiological health were collected at the nurse visit, but not all of the original interviewees agreed to participate in the nurse visit. Models of measured health outcomes (for example forced expiratory volume, fev) could be estimated on the sample of individuals who responded to the nurse visit. But the selection problem means that it may not be possible to make inferences about the determinants of health outcome in the population as a whole. If there are systematic differences between the type of individuals who respond and those who do not, analysts are faced with a fundamental problem of identification.
