A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution

Mixture distribution refers to the combination of more than one probability distribution. Meanwhile, non-normality of data set may be inevitable and the cause may be as a result of mixed distributions thereby renders parametric tests ineffective. Montecarlo experiment was performed 5000 times under...

Full description

Saved in:
Bibliographic Details
Main Authors: T. J. Adejumo, A. A. Akomolafe, A. I. Okegbade, S. D. Gbolagade
Format: Article
Language:English
Published: Nigerian Society of Physical Sciences 2022-12-01
Series:African Scientific Reports
Subjects:
Online Access:https://asr.nsps.org.ng/index.php/asr/article/view/35
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849420983077699584
author T. J. Adejumo
A. A. Akomolafe
A. I. Okegbade
S. D. Gbolagade
author_facet T. J. Adejumo
A. A. Akomolafe
A. I. Okegbade
S. D. Gbolagade
author_sort T. J. Adejumo
collection DOAJ
description Mixture distribution refers to the combination of more than one probability distribution. Meanwhile, non-normality of data set may be inevitable and the cause may be as a result of mixed distributions thereby renders parametric tests ineffective. Montecarlo experiment was performed 5000 times under twelve sample sizes where data were generated from Gaussian and Cauchy distributions using R-statistical packages. At three commonly used alpha levels (0.1, 0.05 and 0.01), the robustness of the test statistics (Rank transformation t-test, Wilcoxon sign test (Distribution and Asymptotic), Signed rank test (Distribution and Asymptotic) and Trimmed t-test) were examined. When the type I error rate of a statistic approximately equal to the true error rate then the statistic is considered robust. At 0.1 and 0.05, Rank transformation t-test, Wilcoxon sign test (distribution) and Trimmed t-test in this order are robust. Meanwhile, at 0.01 Rank transformation andWilcoxon sign test (distribution) were identified to be robust. Also, further counts at all levels of significance revealed that the Rank transformation test is robust and thereby recommended when data comes from a mixed distribution. Hence, this study has been able to identify test statistics that are robust when data comes from a mixed distribution in one sample problem.
format Article
id doaj-art-2a0dd3fea8c24136bfda5a678b849f42
institution Kabale University
issn 2955-1625
2955-1617
language English
publishDate 2022-12-01
publisher Nigerian Society of Physical Sciences
record_format Article
series African Scientific Reports
spelling doaj-art-2a0dd3fea8c24136bfda5a678b849f422025-08-20T03:31:34ZengNigerian Society of Physical SciencesAfrican Scientific Reports2955-16252955-16172022-12-011310.46481/asr.2022.1.3.3535A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture DistributionT. J. AdejumoA. A. AkomolafeA. I. OkegbadeS. D. Gbolagade Mixture distribution refers to the combination of more than one probability distribution. Meanwhile, non-normality of data set may be inevitable and the cause may be as a result of mixed distributions thereby renders parametric tests ineffective. Montecarlo experiment was performed 5000 times under twelve sample sizes where data were generated from Gaussian and Cauchy distributions using R-statistical packages. At three commonly used alpha levels (0.1, 0.05 and 0.01), the robustness of the test statistics (Rank transformation t-test, Wilcoxon sign test (Distribution and Asymptotic), Signed rank test (Distribution and Asymptotic) and Trimmed t-test) were examined. When the type I error rate of a statistic approximately equal to the true error rate then the statistic is considered robust. At 0.1 and 0.05, Rank transformation t-test, Wilcoxon sign test (distribution) and Trimmed t-test in this order are robust. Meanwhile, at 0.01 Rank transformation andWilcoxon sign test (distribution) were identified to be robust. Also, further counts at all levels of significance revealed that the Rank transformation test is robust and thereby recommended when data comes from a mixed distribution. Hence, this study has been able to identify test statistics that are robust when data comes from a mixed distribution in one sample problem. https://asr.nsps.org.ng/index.php/asr/article/view/35Inferential statisticsMixture distributionNon-normalitySimulation study
spellingShingle T. J. Adejumo
A. A. Akomolafe
A. I. Okegbade
S. D. Gbolagade
A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
African Scientific Reports
Inferential statistics
Mixture distribution
Non-normality
Simulation study
title A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
title_full A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
title_fullStr A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
title_full_unstemmed A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
title_short A Simulation Study on Robustness of One Sample Inferential Statistics in Mixture Distribution
title_sort simulation study on robustness of one sample inferential statistics in mixture distribution
topic Inferential statistics
Mixture distribution
Non-normality
Simulation study
url https://asr.nsps.org.ng/index.php/asr/article/view/35
work_keys_str_mv AT tjadejumo asimulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT aaakomolafe asimulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT aiokegbade asimulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT sdgbolagade asimulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT tjadejumo simulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT aaakomolafe simulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT aiokegbade simulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution
AT sdgbolagade simulationstudyonrobustnessofonesampleinferentialstatisticsinmixturedistribution