Suppose a researcher studying the incidence of cardiovascular disease estimates a regression model by including information on each person’s cholesterol level (as an independent explanatory variable). If the researcher fails to also include information on whether the person is a smoker, will the estimated regression coefficient on cholesterol be biased? If so, will the bias be positive or negative?
-
$\begingroup$ Is this for some subject? $\endgroup$ – Glen_b May 11 '15 at 10:52
Are cholesterol level and smoking correlated? If they are not, then your regression coefficient will not be biased. Otherwise you have so called omitted variable bias: http://en.wikipedia.org/wiki/Omitted-variable_bias