Abstract
Computer stimulations have been used to estimate the efficiency, as measured by the statistical power, of various combinations of design and analysis strategies for case-control studies. Conditions under which the various forms of analysis yield consistent relative risk estimators are derived for the general model. The results indicate that the loss of efficiency resulting from the use of a less than optimum design or analysis strategy in many real life situations is small. Practical considerations are of more importance than theoretical statistical ones in deciding upon appropriate strategies. It is concluded that matching is rarely, if ever, justified in most case-control studies of chronic diseases. (Author/publisher)