Abstract:
|
In this paper, we consider a population for which ranks of population units are available. We first identify N units without measurement form this population and determine their ranks. From these N identified units, we construct a simple random sample of size n (n < N) without replacement and measure all of them. We expand this simple random sample to size M (N>M >n-1)) by inserting (M-n) additional units selected at random from the remaining unmeasured N-n units in the population. We now compute the probability distribution of the ranks of measured units in the expanded sample by conditioning on their ranks among N unmeasured units. The final sample contains n measured units and conditional distribution of their ranks in the expanded sample. We construct an estimator for the population mean and show that it has substantial amount of improvement in efficiency over its competitors.
|