Liang, Wei and Dai, Hongsheng and He, Shuyuan (2019) Mean Empirical Likelihood. Computational Statistics and Data Analysis, 138. pp. 155-169. DOI https://doi.org/10.1016/j.csda.2019.04.007
Liang, Wei and Dai, Hongsheng and He, Shuyuan (2019) Mean Empirical Likelihood. Computational Statistics and Data Analysis, 138. pp. 155-169. DOI https://doi.org/10.1016/j.csda.2019.04.007
Liang, Wei and Dai, Hongsheng and He, Shuyuan (2019) Mean Empirical Likelihood. Computational Statistics and Data Analysis, 138. pp. 155-169. DOI https://doi.org/10.1016/j.csda.2019.04.007
Abstract
Empirical likelihood methods are widely used in different settings to construct the confidence regions for parameters which satisfy the moment constraints. However, the empirical likelihood ratio confidence regions may have poor accuracy, especially for small sample sizes and multi-dimensional situations. A novel Mean Empirical Likelihood (MEL) method is proposed. A new pseudo dataset using the means of observation values is constructed to define the empirical likelihood ratio and it is proved that this MEL ratio satisfies Wilks’ theorem. Simulations with different examples are given to assess its finite sample performance, which shows that the confidence regions constructed by Mean Empirical Likelihood are much more accurate than that of the other Empirical Likelihood methods.
Item Type: | Article |
---|---|
Uncontrolled Keywords: | Confidence interval; Empirical likelihood; Exponentially tilted likelihood; Two sample comparison |
Subjects: | Q Science > QA Mathematics |
Divisions: | Faculty of Science and Health Faculty of Science and Health > Mathematics, Statistics and Actuarial Science, School of |
SWORD Depositor: | Unnamed user with email elements@essex.ac.uk |
Depositing User: | Unnamed user with email elements@essex.ac.uk |
Date Deposited: | 16 May 2019 11:11 |
Last Modified: | 30 Oct 2024 16:19 |
URI: | http://repository.essex.ac.uk/id/eprint/24427 |
Available files
Filename: MEL_CSDA_r1.pdf
Licence: Creative Commons: Attribution-Noncommercial-No Derivative Works 3.0