東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Essays on combination of forecasts.

Huang, Huiyu.

FindBook

Google Book

Amazon

博客來

Essays on combination of forecasts.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Essays on combination of forecasts./
作者:	Huang, Huiyu.
面頁冊數:	125 p.
附註:	Adviser: Tae-Hwy Lee.
Contained By:	Dissertation Abstracts International68-06A.
標題:	Economics, Finance. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3270433
ISBN:	9780549089032

Essays on combination of forecasts.
Huang, Huiyu.

Essays on combination of forecasts. - 125 p.

Adviser: Tae-Hwy Lee.

Thesis (Ph.D.)--University of California, Riverside, 2007.

This dissertation consists of four chapters addressing analytically and empirically the merits of combination of forecasts (CF), which combines individual forecasts generated from models each incorporating a part of the whole available information set. Chapter 1 provides further understandings on the out-of-sample success of CF methods, equally weighted CF in particular, in comparison with combination of information (CI) method that pools all available information into one model to generate an ultimate forecast. Analysis is conducted in a weakly stationary environment and then the comparison between CF and CI is applied to equity premium prediction where CF with (close to) equal weights is found to perform generally the best and dominate all CI schemes. Chapter 2 expands the analysis to a nonstationary environment where the data generation process experiences extraneous structural breaks in forecasting period and the CI model is misspecified. Analytical and simulation results show that CF with equal weights is more successful in cases of low signal to noise ratio, strong correlations among predictors that exhibit low persistence, and/or comparably larger breaks. Based on these findings, Chapter 3 proposes a new CF method, the CF using Nelson-Siegel factorizing framework, that is shown to extract the entire information in the yield curve more efficiently in forecasting output growth and inflation than CI methods as well as methods using principal component approach to extract factors from the yield curve. Chapter 4 compares different CF and CI methods that incorporate intraday high frequency information parsimoniously, including subsample averaging, MIDAS, and factor models with principal component approach, in predicting quantiles, mean, and realized volatility of daily S&P 500 returns. In quantile and mean forecasts, generally combining forecasts with simple weighting schemes, subsample averaging in particular, is found to be superior to others.

ISBN: 9780549089032Subjects--Topical Terms:

626650
Economics, Finance.

Essays on combination of forecasts.
LDR:02825nam 2200277 a 45 001 962184
005 20110830
008 110831s2007 ||||||||||||||||| ||eng d
020 $a 9780549089032
035 $a (UMI)AAI3270433
035 $a AAI3270433
040 $a UMI $c UMI
100 1 $a Huang, Huiyu. $3 1285228
245 1 0 $a Essays on combination of forecasts.
300 $a 125 p.
500 $a Adviser: Tae-Hwy Lee.
500 $a Source: Dissertation Abstracts International, Volume: 68-06, Section: A, page: 2566.
502 $a Thesis (Ph.D.)--University of California, Riverside, 2007.
520 $a This dissertation consists of four chapters addressing analytically and empirically the merits of combination of forecasts (CF), which combines individual forecasts generated from models each incorporating a part of the whole available information set. Chapter 1 provides further understandings on the out-of-sample success of CF methods, equally weighted CF in particular, in comparison with combination of information (CI) method that pools all available information into one model to generate an ultimate forecast. Analysis is conducted in a weakly stationary environment and then the comparison between CF and CI is applied to equity premium prediction where CF with (close to) equal weights is found to perform generally the best and dominate all CI schemes. Chapter 2 expands the analysis to a nonstationary environment where the data generation process experiences extraneous structural breaks in forecasting period and the CI model is misspecified. Analytical and simulation results show that CF with equal weights is more successful in cases of low signal to noise ratio, strong correlations among predictors that exhibit low persistence, and/or comparably larger breaks. Based on these findings, Chapter 3 proposes a new CF method, the CF using Nelson-Siegel factorizing framework, that is shown to extract the entire information in the yield curve more efficiently in forecasting output growth and inflation than CI methods as well as methods using principal component approach to extract factors from the yield curve. Chapter 4 compares different CF and CI methods that incorporate intraday high frequency information parsimoniously, including subsample averaging, MIDAS, and factor models with principal component approach, in predicting quantiles, mean, and realized volatility of daily S&P 500 returns. In quantile and mean forecasts, generally combining forecasts with simple weighting schemes, subsample averaging in particular, is found to be superior to others.
590 $a School code: 0032.
650 4 $a Economics, Finance. $3 626650
650 4 $a Economics, General. $3 1017424
690 $a 0501
690 $a 0508
710 2 $a University of California, Riverside. $3 791122
773 0 $t Dissertation Abstracts International $g 68-06A.
790 $a 0032
790 1 0 $a Lee, Tae-Hwy, $e advisor
791 $a Ph.D.
792 $a 2007
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3270433