線性模型中強相關(guān)變量的效應(yīng)估計
發(fā)布時間:2025-04-01 01:06
對于線性模型的準(zhǔn)確的統(tǒng)計推斷和預(yù)測,正確且精確的參數(shù)估計是很重要的。在線性模型中,未知參數(shù)的一般估計步驟都是基于高斯-馬爾科夫定理。這一定理確保了最小二乘估計量是具有最小方差的線性無偏估計量。一般線性模型的主要假設(shè)之一是:預(yù)測變量之間是線性獨立的。但是,在線性回歸中預(yù)測變量經(jīng)常出現(xiàn)“幾乎線性相關(guān)”的情況,這就是所謂的預(yù)測變量共線性問題。這一問題的來源在很多線性回歸分析的書中都有很好的記載,主要可以概括為四類:應(yīng)用的數(shù)據(jù)收集方法,模型或者數(shù)據(jù)總體的約束,模型的指定和過定義模型。理解共線性的來源對于數(shù)據(jù)的分析和相應(yīng)模型的解釋是很有幫助的。通常來說,在很多不同的領(lǐng)域(比如無線通信系統(tǒng)和縱向數(shù)據(jù)分析),強相關(guān)的預(yù)測變量是很常見的。例如,天線陣列中具有一定天線間距的兩個信號是相互關(guān)聯(lián)的?v向數(shù)據(jù)分析通常涉及對一個對象進行多次測量。在這種情況下,同一個對象的多個測量值是相關(guān)的變量。當(dāng)用作是預(yù)測變量時,這些具有強相關(guān)性或極強相關(guān)性的變量引起多重共線性。這種多重共線性問題導(dǎo)致這些強相關(guān)預(yù)測變量的無偏估計量具有異常大的方差,甚至錯誤的符號或者很大的絕對值,從而產(chǎn)生誤導(dǎo)性的統(tǒng)計預(yù)測和推斷。關(guān)于診斷多重共線性...
【文章頁數(shù)】:57 頁
【學(xué)位級別】:碩士
【文章目錄】:
Abstract (In Chinese)
Abstract (In English)
Chapter 1 Introduction
1.1 Background
1.2 Literature Review
1.2.1 The Source of Multicollinearity
1.2.2 Multicollinearity Diagnostics
1.2.3 The Impacts of Multicollinearity
1.2.4 The Remedies for Multicollinearity
1.3 Significance
Chapter 2 The Optimal Effect in an Exponential Model
2.1 The Exponential Model
2.2 The Optimal Group Effect
Chapter 3 Numerical Examples
3.1 Generation of the Exponential Correlation Matrix
3.2 Effect Estimations for Fixed Exponential Correlation Matrix
3.3 Effect Estimations for Random Exponential Correlation Matrix
Chapter 4 Neighbourhood of the Optimal Effect Weights
4.1 The Uniform Model
4.2 The Exponential Model
Conclusions (In English)
Conclusions (In Chinese)
References
Acknowledgements
本文編號:4038704
【文章頁數(shù)】:57 頁
【學(xué)位級別】:碩士
【文章目錄】:
Abstract (In Chinese)
Abstract (In English)
Chapter 1 Introduction
1.1 Background
1.2 Literature Review
1.2.1 The Source of Multicollinearity
1.2.2 Multicollinearity Diagnostics
1.2.3 The Impacts of Multicollinearity
1.2.4 The Remedies for Multicollinearity
1.3 Significance
Chapter 2 The Optimal Effect in an Exponential Model
2.1 The Exponential Model
2.2 The Optimal Group Effect
Chapter 3 Numerical Examples
3.1 Generation of the Exponential Correlation Matrix
3.2 Effect Estimations for Fixed Exponential Correlation Matrix
3.3 Effect Estimations for Random Exponential Correlation Matrix
Chapter 4 Neighbourhood of the Optimal Effect Weights
4.1 The Uniform Model
4.2 The Exponential Model
Conclusions (In English)
Conclusions (In Chinese)
References
Acknowledgements
本文編號:4038704
本文鏈接:http://sikaile.net/jingjilunwen/hongguanjingjilunwen/4038704.html
最近更新
教材專著