{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_52698"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_52698","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"VARIABLE SELECTION AND ESTIMATION IN PARTIAL LINEAR MODEL : A VARIATIONAL BAYES APPROACH","abstract":"Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2025","abstract_html":"Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2025","abstract_has_math":false,"creators":["Dutta, Tathagata"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Maiti, Tapabrata","Bhattacharya, Shrijita","Sung, Chih-Li","Bhattacharya, Sudin"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025","date_published":"2025","updated_at":"2026-11-18T18:59:26Z","subjects":["Statistics"],"languages":["eng"],"rights":["Attribution 4.0 International"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52698","local:Dutta_grad.msu_0128D_21668"],"render_values":[{"text":"etd:52698","href":null,"code":true},{"text":"local:Dutta_grad.msu_0128D_21668","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/0kz4-0394","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/0kz4-0394","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_52698","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Maiti, Tapabrata","Bhattacharya, Shrijita","Sung, Chih-Li","Bhattacharya, Sudin"]},{"key":"dc:creator","label":"Author","values":["Dutta, Tathagata"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2025"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Statistics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution 4.0 International"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52698","local:Dutta_grad.msu_0128D_21668","https://doi.org/doi:10.25335/0kz4-0394"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2025","Partial Linear Models and their generalized extensions represent an important class of semiparametric models that combine the strengths of linear regression with the flexibility of nonparametric function estimation. These models are especially appealing because they include both a linear component, which captures straightforward additive effects of covariates, and a nonparametric component, which allows for complex, unknown relationships without imposing rigid parametric assumptions. This balance of interpretability and adaptability has made Partial Linear Models and Generalized Partial Linear Models widely used in many applied fields.Traditionally, estimation of the nonparametric component in these models has relied on spline-based techniques or kernel smoothing methods. While effective in certain contexts, these approaches can be limited in their capacity to capture highly nonlinear or high-dimensional structures. In this work, we propose a more flexible alternative by employing neural networks to approximate the nonparametric function. Neural networks offer rich representational power and can adaptively learn intricate patterns in the data, making them especially well suited to modeling complex dependencies that are difficult to pre-specify with classical smoothers.To address variable selection in both the linear and nonlinear parts of the model, we incorporate a spike-and-slab prior distribution. This prior combines a point mass at zero (the “spike”) with a continuous distribution (the “slab”), enabling the model to automatically identify and shrink irrelevant predictors while retaining important signals. This approach improves interpretability and helps guard against overfitting, particularly in settings where the number of potential covariates is large.Inference in this model is challenging due to the high dimensionality and the presence of neural network parameters. To achieve computational scalability, we develop a variational inference strategy that approximates the true posterior distribution of model parameters by a more tractable family of distributions. Variational inference provides an efficient optimization-based framework for Bayesian estimation, enabling the method to handle large datasets and complex model structures that are beyond the reach of traditional sampling-based approaches.This combination of neural network estimation, spike-and-slab priors for variable selection, and variational inference creates a powerful toolkit for semiparametric modeling in high-dimensional settings. By moving beyond conventional smoothing techniques, our approach opens new possibilities for flexible, interpretable, and scalable estimation in Partial Linear Models and Generalized Partial Linear Models. We believe these contributions will support further research into modern semiparametric methods and broaden their applicability to contemporary scientific problems where both structure and flexibility are essential.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (94 p.) : ill."]},{"key":"dc:title","label":"Title","values":["VARIABLE SELECTION AND ESTIMATION IN PARTIAL LINEAR MODEL : A VARIATIONAL BAYES APPROACH"]}]}],"canonical_facts":{"dc:contributor":["Maiti, Tapabrata","Bhattacharya, Shrijita","Sung, Chih-Li","Bhattacharya, Sudin"],"dc:creator":["Dutta, Tathagata"],"dc:date":["2025"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Statistics - Doctor of Philosophy, 2025","Partial Linear Models and their generalized extensions represent an important class of semiparametric models that combine the strengths of linear regression with the flexibility of nonparametric function estimation. These models are especially appealing because they include both a linear component, which captures straightforward additive effects of covariates, and a nonparametric component, which allows for complex, unknown relationships without imposing rigid parametric assumptions. This balance of interpretability and adaptability has made Partial Linear Models and Generalized Partial Linear Models widely used in many applied fields.Traditionally, estimation of the nonparametric component in these models has relied on spline-based techniques or kernel smoothing methods. While effective in certain contexts, these approaches can be limited in their capacity to capture highly nonlinear or high-dimensional structures. In this work, we propose a more flexible alternative by employing neural networks to approximate the nonparametric function. Neural networks offer rich representational power and can adaptively learn intricate patterns in the data, making them especially well suited to modeling complex dependencies that are difficult to pre-specify with classical smoothers.To address variable selection in both the linear and nonlinear parts of the model, we incorporate a spike-and-slab prior distribution. This prior combines a point mass at zero (the “spike”) with a continuous distribution (the “slab”), enabling the model to automatically identify and shrink irrelevant predictors while retaining important signals. This approach improves interpretability and helps guard against overfitting, particularly in settings where the number of potential covariates is large.Inference in this model is challenging due to the high dimensionality and the presence of neural network parameters. To achieve computational scalability, we develop a variational inference strategy that approximates the true posterior distribution of model parameters by a more tractable family of distributions. Variational inference provides an efficient optimization-based framework for Bayesian estimation, enabling the method to handle large datasets and complex model structures that are beyond the reach of traditional sampling-based approaches.This combination of neural network estimation, spike-and-slab priors for variable selection, and variational inference creates a powerful toolkit for semiparametric modeling in high-dimensional settings. By moving beyond conventional smoothing techniques, our approach opens new possibilities for flexible, interpretable, and scalable estimation in Partial Linear Models and Generalized Partial Linear Models. We believe these contributions will support further research into modern semiparametric methods and broaden their applicability to contemporary scientific problems where both structure and flexibility are essential.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (94 p.) : ill."],"dc:identifier":["etd:52698","local:Dutta_grad.msu_0128D_21668","https://doi.org/doi:10.25335/0kz4-0394"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["Attribution 4.0 International"],"dc:subject":["Statistics"],"dc:title":["VARIABLE SELECTION AND ESTIMATION IN PARTIAL LINEAR MODEL : A VARIATIONAL BAYES APPROACH"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:26Z"}