| dc.contributor | 國際經營與貿易學系 | - |
| dc.creator (作者) | 顏佑銘 | zh-tw |
| dc.date (日期) | 2016 | - |
| dc.date.accessioned | 14-七月-2017 09:16:49 (UTC+8) | - |
| dc.date.available | 14-七月-2017 09:16:49 (UTC+8) | - |
| dc.date.issued (上傳時間) | 14-七月-2017 09:16:49 (UTC+8) | - |
| dc.identifier.uri (URI) | http://nccur.lib.nccu.edu.tw/handle/140.119/111192 | - |
| dc.description.abstract (摘要) | 近似因子模型及由其衍生出來的各種計量方法，目前被廣泛地應用在各種預測及經濟 分析上。究其原因，乃是近似因子模型可以幫助研究者有效地從大量相關變量中提取 對研究有用的訊息。在近似因子模型中，我們通常假設預測因子之間要有一定的共同 性。在這個計畫裏，我們將著重於有效地估計一種近似因子模型，其中的預測因子除 了受到共同性因素的影響之外，另外也受到一些非共同性因素，如不尋常的巨大異常 值的影響。以下我們列出本計畫會從事的工作項目:(1)我們將發展一個可行的計量方 法來估計這種近似因子模型，而該計量法方法將基於以下的假設:預測因子間的非共 同性因素的出現頻率非常的低; (2)在此假設下，我們將提出了一種懲罰最小平方估 計法(penalized least squares) 來同時分解並估計預測因子的共同及非共同性因 素; (3)為了解決這個估計問題，我們將會開發一個有效率及具彈性的演算程序，而 這項工作將有賴於一些最近提出的優化方法; (4)之後我們會經由大量的蒙地卡羅模 擬，來比較我們所提出的方法和傳統的主成分分析法，在有限樣本下，何者比較能有 效地估計這種近似因子模型; (5)最後我們會將我們所提出的方法用於預測重要總體 經濟指標的年成長率及探討隱性變量如何影響橫斷面預期資產收益率。 Approximate factor models and their extensions are widely used in forecasting and economic analysis due to their ability to extracting useful information from a large number of relevant variables. In these models, candidate predictors are typically subject to some common components. In this project, we will focus on e ciently estimating an approximate factor model in which the candidate predic- tors are additionally subject to idiosyncratic large uncommon components such as jumps or outliers. We outline our plan for the project as follows: (1) We will de- velop a viable econometric method to estimate such an approximate factor model. The econometric method will be based on the assumption that occurrences of the uncommon components are rare; (2) Under this assumption, we will propose a penalized least squares estimation procedure to simultaneously disentangle and estimate the common and uncommon components; (3) To solve the estimation problem, we will develop an e cient and exible algorithm, which will rely on some recently developed optimization methods; (4) We will conduct an intensive Monte-Carlo simulation study to compare nite-sample e ciency of the proposed method and traditional PCA method; (5) We also will demonstrate performances of the proposed method with empirical applications on forecasting yearly growths of important macroeconomic indicators and investigating how latent factors a ect cross sectional expected asset returns. | - |
| dc.format.extent | 1005618 bytes | - |
| dc.format.mimetype | application/pdf | - |
| dc.relation (關聯) | 科技部 | - |
| dc.relation (關聯) | 103-2410-H-004-213 | - |
| dc.subject (關鍵詞) | 近似因子模型;主成分分析法;懲罰最小平方估計法;預測;隱性變量 ;Approximate Factor Model; PCA; Penalized Least Squares; Forecast; Latent Factors | - |
| dc.title (題名) | 近似因子模型的有效估計-經由懲罰最小平方法 | zh-TW |
| dc.title.alternative (其他題名) | Efficient Estimations of Approximate Factor Models via Penalized Leases Squares | en-US |
| dc.type (資料類型) | report | - |
