A comparison of a modified Tillé sampling procedure to Poisson sampling

This paper discusses in detail issues dealing with the technical aspects of designing and conducting surveys. It is intended for an audience of survey methodologists.

Since the late 1950s, the probability surveys in the manufacturing sector within the Manufacturing and Construction Division (MCD) had been almost exclusively selected by using Poisson sampling with unit probabilities assigned proportionate to some measure of size. Poisson sampling has the advantage of simplistic variance calculations. Its disadvantage is that the sample size is a random variable, thus adding an additional (and usually positive) component of variance to the survey estimates. In the 1998 survey year, MCD initiated the use of the modified Tillé sampling procedure in some of its surveys. This sampling procedure is used when there is unequal probability of selection and the sample size is fixed. This paper briefly describes this modified procedure and some of its features, and for a variety of dissimilar surveys, itcontrasts variance results obtained using the Tillé procedure to those resulting from the earlier Poisson procedure.