Some new strategies for estimating area level parameters using information from successive surveys

When two successive surveys are conducted on the same population and the characteristics of interest under study are stable over time then the two surveys can be aggregated to enhance the sample sizes for smaller sub-populations. Apart from the aggregation of the two surveys, effective utilization of available auxiliary information can significantly improve the estimates of interest in those sup-populations. In this direction, we propose four strategies for producing estimates at more granular sub-population levels by combining the information contained in the two surveys under direct, synthetic, and composite methods. The performance of the mean estimators under the proposed strategies is evaluated through a bootstrapped study using the Pakistan Demographic Health Surveys (PDHS 2017–18 and PDHS 19-special) at the regional level. For different choices of the parameters involved in the strategies, the best-performing estimators and best-performing strategies are selected. For the majority of regions, Strategy 3 (S3) and for a few regions Strategy 2 (S2) outperform other strategies in terms of mean squared error (MSE) and percentage contribution of bias (PCoB) in MSE. The selected estimator from each strategy is used to obtain estimates of the parameters (totals or means) of reproductive health characteristics in different geographical units of Pakistan. An R Package is established to obtain the estimated sample sizes, estimates of mean along with their root mean square error, and 95% confidence intervals using the suggested strategies. It is recommended to use S2 and S3 when at least two non-zero observations are available in both surveys with suitable auxiliary information on population. However, in case of <2 observation the S4 is recommended conditional availability of the auxiliary information.