Observational studies and controlled experiments have the same goal, inference about treatment effects, but random assignment of treatments is present only in experiments. This chapter reviews the role of randomization in experiments, and so prepares for discussion of observational studies in later chapters. A theory of observational studies must have a clear view of the role of randomization, so it can have an equally clear view of the consequences of its absence. Sections 2.1 and 2.2 give two examples, first a large controlled clinical trial, and then a small but famous example due to Sir Ronald Fisher, who is usually credited with the invention of randomization, which he called the “reasoned basis for inference” in experiments. Later sections discuss the meaning of this phrase, that is, the link between randomization and statistical methods. Most of the material in this chapter is quite old.
KeywordsPartial Order Random Assignment Treatment Assignment Isotonic Function Adjusted Response
Unable to display preview. Download preview PDF.
- Campbell, D. and Stanley, J. (1963). Experimental and Quasi-Experimental Designs for Research. Chicago: Rand McNally.Google Scholar
- Cochran, W.G. (1963). Sampling Techniques. New York: Wiley.Google Scholar
- Fisher, R.A. (1935, 1949). The Design of Experiments. Edinburgh: Oliver & Boyd.Google Scholar
- Hamilton, M. (1979). Choosing a parameter for 2 x 2 table or 2 x 2 x 2 table analysis. American Journal of Epidemiology, 109, 362–375.Google Scholar
- Mantel, N. and Haenszel, W. (1959). Statistical aspects of retrospective studies of disease. Journal of the National Cancer Institute, 22, 719–748.Google Scholar
- Rubin, D.B. (1986). Which ifs have causal answers? Journal of the American Statistical Association, 81, 961–962.Google Scholar
- Tukey, J.W. (1985). Improving crucial randomized experiments—especially in weather modification—by double randomization and rank combination. In: Proceedings of the Berkeley Conference in Honor of Jerzy Neyman and Jack Kiefer (eds., L. Le Cam and R. Olshen), Volume 1, Belmont, CA: Wadsworth, pp. 79–108.Google Scholar