Estimation of a Difference in Finite Populations with Missing Observations

Area of Science:

• Statistics
• Survey Methodology

Background:

• Missing data is a common challenge in statistical analysis, impacting the reliability of estimates.
• Non-response and deliberate cost-saving measures can lead to incomplete datasets.

Purpose of the Study:

• To develop effective sampling strategies for situations with missing observations.
• To address both non-random sampling for cost reduction and non-response issues.

Main Methods:

• Construction of sampling strategies tailored for missing data.
• Development of an unbiased minimum variance estimator for non-random sampling.
• Deduction of an approximation for the unbiased estimator.
• Application of subsampling tactics for unbiased estimation in non-response cases.

Main Results:

• An unbiased minimum variance estimator was achieved for cost-reduced sampling.
• An approximation to this estimator was successfully deduced.
• Unbiased estimation strategies were developed for non-response scenarios using subsampling.

Conclusions:

• The proposed sampling strategies effectively handle missing data in statistical analysis.
• The methods provide efficient estimation techniques for both planned and unplanned data reduction.