Some aspects of the probability sampling technique of controlled selection.

Data for the 1961 universe of nonfederal, short-term general medical hospitals in the United States are used to illustrate application of estimation and variance formulas for controlled selection, the probability sampling technique developed by Goodman and Kish. Some advantages of this sampling technique are discussed, and a controlled selection model is described. Comparisons are made between the variances of multiple stratification and those of controlled selection; the variances of the latter are seen to have two components, a between pattern and a within pattern component, where a pattern may be regarded as a first stage sample specifying the number of units to be drawn from each control cell. Three strategies for approximating the sampling variance are investigated. Each is seen to overestimate the per hospital estimate of bed capacity, admissions, and inpatient days, but to underestimate slightly the variance of average length of stay. The report that follows is believed to describe the first experience in programming controlled selection for an electronic computer. Steps in the preparatory work, whether for the manual or for the computer operation, are outlined. Some suggestions for continued program development are made.