Top

Previous

Document: 1 of 20

Next

Bottom

Order

Download

Title: A Bayesian Approach to Hypothesis Testing of the Kandahar Province Opinion Polls
Author(s): Eles, P.; Vincent, E.; Kalantzis, E.
Corporate author(s): Defence R&D Canada - Center for Operational Research and Analysis, Ottawa ONT (CAN)

Abstract: This technical memorandum reports on a Bayesian statistical hypothesis test that was developed to aid in the analysis of the Kandahar Province Opinion Polls conducted in support of Canada’s mission in Afghanistan. The test described herein was used by analysts to determine whether changes in public opinion between successive waves of polling or differences between geographic regions were statistically significant. For a number of practical reasons, multi-stage stratified clustered sampling was employed in the Kandahar polls rather than simple random sampling. The test was developed to incorporate corrections for the effects of non-simple random sampling as well as to handle small sample sizes. This paper demonstrates the validity of the test through simulation and by comparison with a conventional test in the regime where the two are expected to converge.
Il y a un résumé français ici.