Eylem Seç
Topics in Joint Estimation of Vector Autoregressive Models
Başlık:
Topics in Joint Estimation of Vector Autoregressive Models
Yazar:
Skripnikov, Andrey, author.
ISBN:
9780438122185
Yazar Ek Girişi:
Fiziksel Tanımlama:
1 electronic resource (87 pages)
Genel Not:
Source: Dissertation Abstracts International, Volume: 79-11(E), Section: B.
Advisors: George Michailidis.
Özet:
In this work we develop various frameworks for performing joint regularized estimation of multiple related high-dimensional time series. The general topic of estimating high-dimensional time series data has received a lot of attention over the years, with applications spanning from gene expression data to stock market returns, macro-econometric time series to brain fMRI measurements, to name a few. One of the most popular operational tools for such data are Vector Autoregressive models (VAR), the focal point of this work. This class of models allows to account for both the temporal dynamics of the variables under consideration and the contemporaneous dependence between them. The main characteristic of high dimensional problems is the excessive amount of unknown model parameters compared to the number of available observations, which renders most classical estimation methods infeasible. An example would be brain fMRI time series data collected over thousands of brain voxels (parameters) during a limited number of time points (observations). For that, we introduce regularized estimation via various Lasso procedures. In many cases one also encounters multiple related high-dimensional time series, e.g. brain fMRI measurements for patients with the same mental disease (ADHD, Autism, Alzheimer's etc). You expect to see certain features that are common for all subjects, and therefore would want to use the data from all the patients to improve the estimation procedure for each time series. We introduce multiple joint modeling frameworks that help to account for that assumed similarity between the related subjects. Those frameworks will leverage various penalties in order to enforce similarity of resulting estimates. In particular, Chapter 2 will implement fused Lasso penalty for econometric time series of multiple US states with similar economical infrastructure and industrial characteristics, while Chapter 3 will employ group Lasso penalty for the aforementioned application to brain fMRI time series data for groups of patients. While both Chapters 2 and 3 will provide simulation study results showing the superior performance of the joint method over classical estimation approaches, neither provides a well-defined inferential procedure to test for the temporal effect significance. Chapter 4 will contain a hypothesis testing procedure for group effects after performing regularized estimation for each member of the group. 10.
Notlar:
School code: 0070
Tüzel Kişi Ek Girişi:
Mevcut:*
Yer Numarası | Demirbaş Numarası | Shelf Location | Lokasyon / Statüsü / İade Tarihi |
---|---|---|---|
XX(696647.1) | 696647-1001 | Proquest E-Tez Koleksiyonu | Arıyor... |
On Order
Liste seç
Bunu varsayılan liste yap.
Öğeler başarıyla eklendi
Öğeler eklenirken hata oldu. Lütfen tekrar deneyiniz.
:
Select An Item
Data usage warning: You will receive one text message for each title you selected.
Standard text messaging rates apply.