{"id":{"repo_id":"buffalo","oai_identifier":"oai:ubir.buffalo.edu:10477/83840"},"canonical_url":"https://search.dev.ndltd.org/etd/buffalo/oai:ubir.buffalo.edu:10477/83840","repository":{"repo_id":"buffalo","name":"Buffalo","base_url":"https://ubir.buffalo.edu/oai/request"},"display":{"title":"Contributions to the Theory of Statistical Distances with Applications to Safety of Medical Products","abstract":"Ph.D.","abstract_html":"Ph.D.","abstract_has_math":false,"creators":["Ding, Yuxin"],"institution":"State University of New York at Buffalo","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Markatou, Marianthi","Biostatistics"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2022,"date_issued":"2022-06-17T19:54:39Z","date_published":"2022-06-17T19:54:39Z","updated_at":"2026-07-27T19:05:28Z","subjects":["biostatistics","statistics","pharmaceutical sciences"],"languages":["eng"],"rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10477/83840","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Markatou, Marianthi","Biostatistics"]},{"key":"dc:creator","label":"Author","values":["Ding, Yuxin"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2022-06-17T19:54:39Z","2020"]},{"key":"dc:publisher","label":"Institution","values":["State University of New York at Buffalo"]},{"key":"dc:type","label":"Dc Type","values":["Text","Dissertation"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["biostatistics","statistics","pharmaceutical sciences"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/10477/83840"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Ph.D.","Safety of medical products presents a serious concern worldwide. Surveillance systems for continual monitoring of the safety of post market medical products, such as spontaneous reporting systems and pharmacoepidemiology databases, have been established in many countries. In this thesis, we take a critical look at the currently used methods for adverse event identification with an eye towards understanding their fundamental weaknesses, as well as the origins of these weaknesses. We then introduce the pattern discovery framework proposed by Markatou and Ball (2014), and discuss its key components. This framework highlights the need for constructing density estimation methods for mixed-scale data, i.e. interval and nominal/categorical scale data. Suggested in the literature kernel density estimation methods for discrete random variables have a number of drawbacks. For example, the kernels used present difficulties in computation as well as implementation, issues that impact the quality of the probability mass function estimator particularly in the presence of sparse data. Furthermore, they have either very limited or no inferential capacity. We propose a statistical definition of a new class of kernels: the class of diffusion kernels. Diffusion kernels contain tuning parameters that are functions of their degrees of freedom; they allow easy computation and are particularly appropriate for demonstrating the usefulness of product kernels in higher dimensions and when data have both, interval and nominal/categorical scale. We illustrate the wide inferential capacity of diffusion kernels through the use of one important element of this class, the Poisson kernel, in constructing test statistics for testing uniformity on the hyper-sphere. Our proposed Poisson kernel-based tests either outperform all other tests in testing uniformity or they are at least as competitive with existing tests in the literature. The definition of diffusion kernel class and identification of members appropriate for use given the scale of the data constitutes the first step in being able to create high performance solutions for the thorny problem of safety of medical products.","**To request an accessible version of the file(s) associated with this item, contact library@buffalo.edu. Please include the item's persistent URL [http://hdl.handle.net/. . .] in your request.**"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Contributions to the Theory of Statistical Distances with Applications to Safety of Medical Products"]}]}],"canonical_facts":{"dc:contributor":["Markatou, Marianthi","Biostatistics"],"dc:creator":["Ding, Yuxin"],"dc:date":["2022-06-17T19:54:39Z","2020"],"dc:description":["Ph.D.","Safety of medical products presents a serious concern worldwide. Surveillance systems for continual monitoring of the safety of post market medical products, such as spontaneous reporting systems and pharmacoepidemiology databases, have been established in many countries. In this thesis, we take a critical look at the currently used methods for adverse event identification with an eye towards understanding their fundamental weaknesses, as well as the origins of these weaknesses. We then introduce the pattern discovery framework proposed by Markatou and Ball (2014), and discuss its key components. This framework highlights the need for constructing density estimation methods for mixed-scale data, i.e. interval and nominal/categorical scale data. Suggested in the literature kernel density estimation methods for discrete random variables have a number of drawbacks. For example, the kernels used present difficulties in computation as well as implementation, issues that impact the quality of the probability mass function estimator particularly in the presence of sparse data. Furthermore, they have either very limited or no inferential capacity. We propose a statistical definition of a new class of kernels: the class of diffusion kernels. Diffusion kernels contain tuning parameters that are functions of their degrees of freedom; they allow easy computation and are particularly appropriate for demonstrating the usefulness of product kernels in higher dimensions and when data have both, interval and nominal/categorical scale. We illustrate the wide inferential capacity of diffusion kernels through the use of one important element of this class, the Poisson kernel, in constructing test statistics for testing uniformity on the hyper-sphere. Our proposed Poisson kernel-based tests either outperform all other tests in testing uniformity or they are at least as competitive with existing tests in the literature. The definition of diffusion kernel class and identification of members appropriate for use given the scale of the data constitutes the first step in being able to create high performance solutions for the thorny problem of safety of medical products.","**To request an accessible version of the file(s) associated with this item, contact library@buffalo.edu. Please include the item's persistent URL [http://hdl.handle.net/. . .] in your request.**"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/10477/83840"],"dc:language":["eng"],"dc:publisher":["State University of New York at Buffalo"],"dc:rights":["Users of works found in University at Buffalo Institutional Repository (UBIR) are responsible for identifying and contacting the copyright owner for permission to reuse. University at Buffalo Libraries do not manage rights for copyright-protected works and cannot assist with permissions.","Copyright retained by author."],"dc:subject":["biostatistics","statistics","pharmaceutical sciences"],"dc:title":["Contributions to the Theory of Statistical Distances with Applications to Safety of Medical Products"],"dc:type":["Text","Dissertation"]},"updated_at":"2026-07-27T19:05:28Z"}