{"id":{"repo_id":"mcmaster","oai_identifier":"oai:macsphere.mcmaster.ca:11375/33854"},"canonical_url":"https://search.ndltd.org/etd/mcmaster/oai:macsphere.mcmaster.ca:11375/33854","repository":{"repo_id":"mcmaster","name":"McMaster University","base_url":"https://macsphere.mcmaster.ca/server/oai/request"},"display":{"title":"Bayesian Nonparametric Mixture Models for Robust Inference Under Non-Ignorable Missingness","abstract":"Longitudinal studies are often prone to missingness due to participant dropout, loss to follow-up, or even death. When the probability that a value is missing depends on both observed and unobserved values, the data are said to be missing not at random (MNAR), indicating an informative missingness mechanism. This gives rise to the problem of non-ignorable dropout. Some methods historically used to model non-ignorable non-response rely on unverifiable assumptions about the relationship between the outcome and dropout. This issue can be further compounded when there are potential outliers in the model. A Dirichlet process mixture model of contaminated Gaussian distributions is proposed. In this thesis, two extensions of Dirichlet process mixture models are developed for modelling dropout while accounting for outliers and atypical observations. Firstly, we propose a Dirichlet process mixture model with contaminated Gaussian kernels. The model aims to relax parametric assumptions while accounting for outliers in the study. We then develop a Dirichlet process mixture model using the contaminated shifted asymmetric Laplace distribution to accommodate contamination, skewness, and scale parameters. A modified MCMC Gibbs sampler is developed and used for estimation. Analyses are performed on a publicly available HIV dataset used by \\cite{moore2020dirichlet} and on a simulation study.","abstract_html":"Longitudinal studies are often prone to missingness due to participant dropout, loss to follow-up, or even death. When the probability that a value is missing depends on both observed and unobserved values, the data are said to be missing not at random (MNAR), indicating an informative missingness mechanism. This gives rise to the problem of non-ignorable dropout. Some methods historically used to model non-ignorable non-response rely on unverifiable assumptions about the relationship between the outcome and dropout. This issue can be further compounded when there are potential outliers in the model. A Dirichlet process mixture model of contaminated Gaussian distributions is proposed. In this thesis, two extensions of Dirichlet process mixture models are developed for modelling dropout while accounting for outliers and atypical observations. Firstly, we propose a Dirichlet process mixture model with contaminated Gaussian kernels. The model aims to relax parametric assumptions while accounting for outliers in the study. We then develop a Dirichlet process mixture model using the contaminated shifted asymmetric Laplace distribution to accommodate contamination, skewness, and scale parameters. A modified MCMC Gibbs sampler is developed and used for estimation. Analyses are performed on a publicly available HIV dataset used by \\cite{moore2020dirichlet} and on a simulation study.","abstract_has_math":false,"creators":["Sowu, Elorm Afi"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":"Mathematics and Statistics","school":null,"contributors":[],"advisors":["McNicholas , Paul"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T19:37:42Z","subjects":[],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://doi.org/10.71548/25624"],"render_values":[{"text":"https://doi.org/10.71548/25624","href":"https://doi.org/10.71548/25624","code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/11375/33854","outbound_label":"Handle","outbound_source":"dc:identifier.uri","record_url":"https://hdl.handle.net/11375/33854","record_label":"Repository record","record_kind":"handle","record_source":"dc:identifier.uri","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://macsphere.mcmaster.ca/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Amacsphere.mcmaster.ca%3A11375%2F33854","prefix":"dim"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["McNicholas , Paul"]},{"key":"dc:contributor.department","label":"Department","values":["Mathematics and Statistics"]},{"key":"dc:creator","label":"Author","values":["Sowu, Elorm Afi"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-10-15T12:21:29Z"]},{"key":"dc:date.issued","label":"Date","values":["2026"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/11375/33854","https://doi.org/10.71548/25624"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Longitudinal studies are often prone to missingness due to participant dropout, loss to follow-up, or even death. When the probability that a value is missing depends on both observed and unobserved values, the data are said to be missing not at random (MNAR), indicating an informative missingness mechanism. This gives rise to the problem of non-ignorable dropout. Some methods historically used to model non-ignorable non-response rely on unverifiable assumptions about the relationship between the outcome and dropout. This issue can be further compounded when there are potential outliers in the model. A Dirichlet process mixture model of contaminated Gaussian distributions is proposed. In this thesis, two extensions of Dirichlet process mixture models are developed for modelling dropout while accounting for outliers and atypical observations. Firstly, we propose a Dirichlet process mixture model with contaminated Gaussian kernels. The model aims to relax parametric assumptions while accounting for outliers in the study. We then develop a Dirichlet process mixture model using the contaminated shifted asymmetric Laplace distribution to accommodate contamination, skewness, and scale parameters. A modified MCMC Gibbs sampler is developed and used for estimation. Analyses are performed on a publicly available HIV dataset used by \\cite{moore2020dirichlet} and on a simulation study."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Doctor of Philosophy (PhD)"]},{"key":"dc:title","label":"Title","values":["Bayesian Nonparametric Mixture Models for Robust Inference Under Non-Ignorable Missingness"]},{"key":"others:access_status","label":"Others Access Status","values":["embargo"]}]}],"canonical_facts":{"dc:contributor.advisor":["McNicholas , Paul"],"dc:contributor.department":["Mathematics and Statistics"],"dc:creator":["Sowu, Elorm Afi"],"dc:date.accessioned":["2026-10-15T12:21:29Z"],"dc:date.issued":["2026"],"dc:description.abstract":["Longitudinal studies are often prone to missingness due to participant dropout, loss to follow-up, or even death. When the probability that a value is missing depends on both observed and unobserved values, the data are said to be missing not at random (MNAR), indicating an informative missingness mechanism. This gives rise to the problem of non-ignorable dropout. Some methods historically used to model non-ignorable non-response rely on unverifiable assumptions about the relationship between the outcome and dropout. This issue can be further compounded when there are potential outliers in the model. A Dirichlet process mixture model of contaminated Gaussian distributions is proposed. In this thesis, two extensions of Dirichlet process mixture models are developed for modelling dropout while accounting for outliers and atypical observations. Firstly, we propose a Dirichlet process mixture model with contaminated Gaussian kernels. The model aims to relax parametric assumptions while accounting for outliers in the study. We then develop a Dirichlet process mixture model using the contaminated shifted asymmetric Laplace distribution to accommodate contamination, skewness, and scale parameters. A modified MCMC Gibbs sampler is developed and used for estimation. Analyses are performed on a publicly available HIV dataset used by \\cite{moore2020dirichlet} and on a simulation study."],"dc:description.degree":["Doctor of Philosophy (PhD)"],"dc:identifier.uri":["https://hdl.handle.net/11375/33854","https://doi.org/10.71548/25624"],"dc:language.iso":["en"],"dc:title":["Bayesian Nonparametric Mixture Models for Robust Inference Under Non-Ignorable Missingness"],"dc:type":["Thesis"],"others:access_status":["embargo"]},"updated_at":"2026-11-18T19:37:42Z"}