Abstract
The problem of estimating the finite population mean on the samples selected over two occasions has been considered, when there is non-response (i) on both the occasions, (ii) only on the second occasion for both matched and unmatched portions of the sample, and (iii) only on the second occasion for unmatched portion of the sample. For the case when two auxiliary variables are positively and negatively correlated with the study variable, a double sampling ratio-cum-product estimate from the matched portion of the sample is presented. Expressions for optimum matching fraction and of the combined estimate have been derived. The gain in efficiency, the sample sizes and the cost function of the combined estimators over the direct estimate using no information gathered on the first occasion is computed. The comparison between the proposed strategy with other estimators is also carried out. An empirical study is also included for illustration.