Robustness of marginal likelihood based tests for random regression coefficients

Maxwell Leslie King, Shahidur Rahman

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Over the last few decades there has been a growing literature on diagnostic tests of regression disturbances. Tests that have been constructed based on marginal likelihood methods have been found to do well when the disturbances are normally distributed. This paper investigates the small-sample size and power properties of marginal likelihood based tests when testing for random regression coefficients in the presence of first-order autoregressive disturbances. We find test sizes are less robust to non-normality as the sample size increases and the relative power performance of the various tests hardly changes as non-normality is introduced. Consequently marginal likelihood score based tests typically have the best power properties with a particular approximate point optimal invariant test providing some exceptions.
Original languageEnglish
Pages (from-to)205 - 220
Number of pages16
JournalModel Assisted Statistics and Applications
Volume10
Issue number3
DOIs
Publication statusPublished - 2015

Cite this

@article{5900ef8a30d5477295c49e8fad7948c5,
title = "Robustness of marginal likelihood based tests for random regression coefficients",
abstract = "Over the last few decades there has been a growing literature on diagnostic tests of regression disturbances. Tests that have been constructed based on marginal likelihood methods have been found to do well when the disturbances are normally distributed. This paper investigates the small-sample size and power properties of marginal likelihood based tests when testing for random regression coefficients in the presence of first-order autoregressive disturbances. We find test sizes are less robust to non-normality as the sample size increases and the relative power performance of the various tests hardly changes as non-normality is introduced. Consequently marginal likelihood score based tests typically have the best power properties with a particular approximate point optimal invariant test providing some exceptions.",
author = "King, {Maxwell Leslie} and Shahidur Rahman",
year = "2015",
doi = "10.3233/MAS-150325",
language = "English",
volume = "10",
pages = "205 -- 220",
journal = "Model Assisted Statistics and Applications",
issn = "1574-1699",
publisher = "IOS Press",
number = "3",

}

Robustness of marginal likelihood based tests for random regression coefficients. / King, Maxwell Leslie; Rahman, Shahidur.

In: Model Assisted Statistics and Applications, Vol. 10, No. 3, 2015, p. 205 - 220.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Robustness of marginal likelihood based tests for random regression coefficients

AU - King, Maxwell Leslie

AU - Rahman, Shahidur

PY - 2015

Y1 - 2015

N2 - Over the last few decades there has been a growing literature on diagnostic tests of regression disturbances. Tests that have been constructed based on marginal likelihood methods have been found to do well when the disturbances are normally distributed. This paper investigates the small-sample size and power properties of marginal likelihood based tests when testing for random regression coefficients in the presence of first-order autoregressive disturbances. We find test sizes are less robust to non-normality as the sample size increases and the relative power performance of the various tests hardly changes as non-normality is introduced. Consequently marginal likelihood score based tests typically have the best power properties with a particular approximate point optimal invariant test providing some exceptions.

AB - Over the last few decades there has been a growing literature on diagnostic tests of regression disturbances. Tests that have been constructed based on marginal likelihood methods have been found to do well when the disturbances are normally distributed. This paper investigates the small-sample size and power properties of marginal likelihood based tests when testing for random regression coefficients in the presence of first-order autoregressive disturbances. We find test sizes are less robust to non-normality as the sample size increases and the relative power performance of the various tests hardly changes as non-normality is introduced. Consequently marginal likelihood score based tests typically have the best power properties with a particular approximate point optimal invariant test providing some exceptions.

U2 - 10.3233/MAS-150325

DO - 10.3233/MAS-150325

M3 - Article

VL - 10

SP - 205

EP - 220

JO - Model Assisted Statistics and Applications

JF - Model Assisted Statistics and Applications

SN - 1574-1699

IS - 3

ER -