Martin Hyllienmark, Statistiska institutionen

SEMINARIUM
Datum: onsdag 10 juni 2026
Tid: 13:00 - 14:00
Plats: Campus Albano, lärosal 26, hus 4, plan 2

Joint sampling designs and the Joint Difference Estimator

Abstract

In addition to auxiliary information, some estimators also utilize the information from consecutive time periods to improve precision in the estimates. A common drawback of such estimators is the need of tailor-made designs which ensure a certain overlap between samples between the time periods. In this paper, we propose a novel approach to incorporate information from previous study periods in the estimator to reduce variance. We propose a theory called joint sampling design which entails the first and second order inclusion probabilities over several consecutive time periods. The proposed Joint Difference Estimator utilizes these inclusion probabilities combined with a difference estimator. We show that the proposed estimator is unbiased, has a closed-form variance expression and an unbiased variance estimator. It is also design-independent which makes it more flexible compared to e.g. composite estimators which requires an overlapping design. With numerical examples we illustrate that the estimator is competitive compared to well established estimators and some recently developed composite estimators. The main finding of the paper is that the proposed estimator can substantially reduce the total sample size over time by unequal allocation over a set number of time periods.


Senast uppdaterad: 2026-06-04

Sidansvarig: Statistiska institutionen