Self-Selection bias is a long-standing problem in Epidemiology. The proposed research will apply new self-selection bias correction techniques from econometrics to two epidemiological questions: (1) What is the effect of excessive alcohol consumption on blood pressure? and (2) What is the effect of schooling on blood pressure? While the suggested econometric solutions have been widely used to address a variety of economic, psychological, sociological and policy questions, the proposed study appears to offer the first epidemiological applications. Data for both applications will be drawn from the National Health and Nutrition Examination Survey I, 1971-1975, (NHANES). (1) Excessive alcohol consumption and hypertension have been found to be strongly correlated in a number of studies. The correlation is generally thought to indicate that excessive alcohol use causes high blood pressure. What statisticians refer to as the """"""""constitutional or self-selection hypotheses"""""""" (COSSH) may also explain the correlation, however. Heredity may predispose individuals to high blood pressure and/or excessive alcohol use. If hereditary factors cluster together in the same persons, we would expect to observe a correlation between hypertension and excessive alcohol use, but this correlation would not imply causation; it would only reflect genetic clustering. The causation hypothesis is nevertheless difficult to dismiss entirely. A handful of studies have provided evidence that, on average, when excessive drinkers cut their consumption to zero for an extended number of months their blood pressures fall. The real question then is to determine what part of the observed correlation reflects the causal relation and what part reflects unobserved genetic factors. (2) Studies have found an inverse relation between an individual's years of schooling completed and blood pressure, other things equal. Explanations for the correlation differ. Some have argued for a direct effect whereby schooling enhances the """"""""production of health."""""""" Others assert that one or more """"""""third variables"""""""" such as unobserved genetic or personality factors affect both blood pressure and schooling the same direction; this argument is essentially a self-selection explanation. The second application will test if evidence from the NHANESI data provide more support for the """"""""production"""""""" of health argument or the self-selection argument. If evidence for self-selection is obtained, policy recommendations to increase funding on education to improve health are weakened.
| Leigh, J P (1988) Assessing the importance of an independent variable in multiple regression: is stepwise unwise? J Clin Epidemiol 41:669-77 |
