This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Econ 508: Correlation and Regression Juan Fung MSPE February 11, 2011 Juan Fung (MSPE) Econ 508: Correlation and Regression February 11, 2011 1 / 17 Introduction: the purpose of correlation and regression analysis Suppose you observe n realizations of two random variables, X and Y . Example For example, let X = parents’ income (in $1000), Y = student GPA. X Y X Y 21 4.0 12 3.0 15 3.0 18 3.5 15 3.5 6 2.5 9 2.0 12 2.5 Using this data, what questions might you ask regarding X and Y ? 1. Is there a relationship between X and Y ? 2. If so, what is the form of the relationship? For example, do you expect that an increase in parent income would be associated with a proportional increase in student GPA? Juan Fung (MSPE) Econ 508: Correlation and Regression February 11, 2011 2 / 17 Correlation Correlation analysis Recall the correlation coefficient, ρ = Cov [ X , Y ] p Var [ X ] Var [ Y ] ≡ σ XY σ X σ Y If X and Y are independent, then ρ = 0 (since Cov [ X , Y ] ≡ σ XY = 0). Equivalently, if ρ 6 = 0 , then X and Y are not independent . Even more, they share a linear relationship . The direction and strength of this relationship is determined by the sign and magnitude of ρ . The goal of correlation analysis is to identify whether a linear relationship exists, and if it does, its direction and strength. Juan Fung (MSPE) Econ 508: Correlation and Regression February 11, 2011 3 / 17 Correlation Recall example of GPA and parent income. An excel scatter plot suggests positive correlation: But we can’t say anything precise about their relationship. 1. First, ρ is unobservable: we need population moments. 2. Even if we knew they shared a strong positive relationship, ρ tells us nothing about how one variable (say X ) influences another (say Y ). Juan Fung (MSPE) Econ 508: Correlation and Regression February 11, 2011 4 / 17 Correlation ρ is unobservable Luckily, the MLE for ρ (try deriving) has an intuitive formulation, r = S XY √ S XX S YY , where S XX = X i ( X i X ) 2 = ( n 1) d Var ( X ) S YY = X i ( Y i...
View
Full
Document
This note was uploaded on 05/23/2011 for the course ECON 508 taught by Professor Staff during the Spring '08 term at University of Illinois, Urbana Champaign.
 Spring '08
 Staff

Click to edit the document details