Best Linear Unbiased Estimators for Repeated Surveys

Summary Recent work on the problem of obtaining best linear unbiased estimators from a sample survey which is repeated on several occasions has centred on the effects of extending the model assumptions to allow for stochastic variation in the parameters being estimated. In this paper a unified approach to the problem is given using least squares theory. The results of Blight and Scott (1973) are extended, and the relationships between their results, those of Scott and Smith (1974) and the classical approach of Patterson (1950), or its generalization by Gurney and Daly (1965), are clarified. The problems involved in putting the time series estimators into practice are examined, with particular reference to the role of stationarity.