LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION
| Main Author: | |
|---|---|
| Language: | English |
| Published: |
Case Western Reserve University School of Graduate Studies / OhioLINK
2007
|
| Subjects: | |
| Online Access: | http://rave.ohiolink.edu/etdc/view?acc_num=case1164135815 |
| id |
ndltd-OhioLink-oai-etd.ohiolink.edu-case1164135815 |
|---|---|
| record_format |
oai_dc |
| spelling |
ndltd-OhioLink-oai-etd.ohiolink.edu-case11641358152021-08-03T05:32:18Z LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION Xing, Guan LASSO This dissertation includes two parts. The first part describes a new estimation method for finite mixture models. Different from traditional methods, we borrow ideas from the variable selection approaches in linear models. After generating the pseudo-response from a saturated mixture model and constructing predictors using candidate component densities, we transform the mixture model density estimation problem to a variable selection problem in linear models. Using a variant of the LASSO constraint approach, we can do component number selection and parameter estimation simultaneously for finite mixture models. The performance of this method is illustrated with simulated data and some well-known real data sets. In the second part, we deal with the estimation with the contaminated data. Traditional Bayesian approaches use the variance-inflation model or the mean-shift model. We extend the Bayesian contaminated model to the general case without assuming a specific distribution for the potential outliers, but using a constructed reference population to draw random samples. With the proposed latent indicator variables for each observation, we construct a Bayesian hierarchical model and use Gibbs sampler to draw posterior samples. The parameter inference based on the Gibbs samples is robust, and a series of simulations and classicreal data analysis indicate the better performance of our methods than other approaches in linear models, generalized linear models and density estimation. 2007 English text Case Western Reserve University School of Graduate Studies / OhioLINK http://rave.ohiolink.edu/etdc/view?acc_num=case1164135815 http://rave.ohiolink.edu/etdc/view?acc_num=case1164135815 unrestricted 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 |
LASSO |
| spellingShingle |
LASSO Xing, Guan LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| author |
Xing, Guan |
| author_facet |
Xing, Guan |
| author_sort |
Xing, Guan |
| title |
LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| title_short |
LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| title_full |
LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| title_fullStr |
LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| title_full_unstemmed |
LASSOING MIXTURES AND BAYESIAN ROBUST ESTIMATION |
| title_sort |
lassoing mixtures and bayesian robust estimation |
| publisher |
Case Western Reserve University School of Graduate Studies / OhioLINK |
| publishDate |
2007 |
| url |
http://rave.ohiolink.edu/etdc/view?acc_num=case1164135815 |
| work_keys_str_mv |
AT xingguan lassoingmixturesandbayesianrobustestimation |
| _version_ |
1719421431405608960 |