Statistical models and algorithms for large data with complex dependence structures

Statistical models and algorithms for large data with complex dependence structures

Bibliographic Details
Main Author: Li, Miaoqi
Language:English
Published: University of Cincinnati / OhioLINK 2020
Subjects:
Statistics
algorithms
complex dependence structures
large data
statistical models
Online Access:http://rave.ohiolink.edu/etdc/view?acc_num=ucin1584015958922068
id ndltd-OhioLink-oai-etd.ohiolink.edu-ucin1584015958922068
record_format oai_dc
spelling ndltd-OhioLink-oai-etd.ohiolink.edu-ucin15840159589220682021-08-03T07:13:57Z Statistical models and algorithms for large data with complex dependence structures Li, Miaoqi Statistics algorithms complex dependence structures large data statistical models In this dissertation, our research interest focuses on developing statistical models and algorithms for large-scale data. Specifically, we are considering data with complex dependence structures. Two types of complicated dependence structures have been addressed. They are large-scale network data and multivariate processes with spatial dependence.Network data possess intricate configurations, which causes conducting thorough investigations on properties and making inferences on large-scale networks to be challenged and even infeasible. To overcome these difficulties, we take advantage of recent developments in randomized numerical linear algebra and derive efficient algorithms to estimate the spatial autocorrelation parameter by approximating log likelihood function of the spatial autoregressive (SAR) model.When studying multivariate processes with spatial dependence, we propose a multivariate fused Gaussian process (MFGP) model that is able to flexibly model multivariate spatial processes and enables efficient computation. The proposed model combines a low-rank component and a multivariate Gaussian Markov random field to jointly depict spatial dependence structure that is potentially large-scale, non-stationary and asymmetric.Compelling experimental results from extensive simulation and real data examples demonstrate empirically that the performance and applications of our proposed models and algorithms are better than many state-of-the-art methodologies based on a variety of criteria. The theoretical properties are explored and consistency results are established. 2020-06-02 English text University of Cincinnati / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=ucin1584015958922068 http://rave.ohiolink.edu/etdc/view?acc_num=ucin1584015958922068 restricted--full text unavailable until 2022-05-20 This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.
collection NDLTD
language English
sources NDLTD
topic Statistics
algorithms
complex dependence structures
large data
statistical models
spellingShingle Statistics
algorithms
complex dependence structures
large data
statistical models
Li, Miaoqi
Statistical models and algorithms for large data with complex dependence structures
author Li, Miaoqi
author_facet Li, Miaoqi
author_sort Li, Miaoqi
title Statistical models and algorithms for large data with complex dependence structures
title_short Statistical models and algorithms for large data with complex dependence structures
title_full Statistical models and algorithms for large data with complex dependence structures
title_fullStr Statistical models and algorithms for large data with complex dependence structures
title_full_unstemmed Statistical models and algorithms for large data with complex dependence structures
title_sort statistical models and algorithms for large data with complex dependence structures
publisher University of Cincinnati / OhioLINK
publishDate 2020
url http://rave.ohiolink.edu/etdc/view?acc_num=ucin1584015958922068
work_keys_str_mv AT limiaoqi statisticalmodelsandalgorithmsforlargedatawithcomplexdependencestructures
_version_ 1719457071880994816

Similar Items