Experimental Design in Epidemiologic Research

¹ Zahnärztliches Institut, Zurich, Switzerland

It is obvious that problems of design of epidemiologic studies vary considerably with the scope and purpose of the study. The important feature of the statistical design is that it has to be planned in accordance with clearly defined principles. The only way by which it is possible to exclude bias in the sample is to draw a random sample from the population under study. In a true random sample, every element of the total population has an equal probability of being included. When the sampling problems differ markedly within the population, stratification or subdivision into subpopulations is [see digure in the PDF file] indicated. The usefulness of stratification and partial recording is examined, and problems of economy and the use of electronic computers in the statistical evaluation are discussed.