|
Abstract:
|
The calibration approach for estimation of finite population parameters consists of: (i) computation of weights that incorporates specified auxiliary information, in a systematic way, and are restrained by calibration equations (ii) the use of these weights to compute linearly weighted estimates of total and other finite population parameters, using various distance functions. In the linearly (weighted) estimators, the weights can depend either on the draw or on the units selected in the sample or on the sample selected or a combination of these. This gives rise to various classes of linear estimators [Horwitz & Thompson (1952), Godambe (1955), Koop (1961, 63), Tikkiwal (1965)]. Bhargava and Tikkiwal (1978) reviewed the earlier work and gave a modified set of classes. This paper obtains improved estimators through application of the calibration approach to different classes of linear estimators using chi-square and entropy distance functions under varying probability sampling design.
|
