This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 1 Hypothesis Testing and Confidence Interval for Simple Linear Regression 1. So far we have discussed the distributions of ) , ( 1 b b or ) ˆ , ˆ ( 1 β β , given by equations (10)(11) [in lecture notes 2]. Note that, we had another estimator for the unknown error variance & = = n i i u n 1 2 2 ˆ 2 1 ˆ σ In this section we discuss the distribution of 2 ˆ σ since these will be used in hypothesis testing as well. First recall the following basic results: Definition of 2 χ (chisquared) distribution: Let n Z Z Z ,........, , 2 1 are IID N(0,1), i.e., each i Z follows a standard normal distribution and they are independent. Then & = n i i Z 1 2 follows a 2 χdistribution with n degrees of freedom (d.f). Note that, under the OLS assumptions we have, ) , ( ~ 2 σ IIDN u i , i.e., ) 1 , ( ~ IIDN u i ¡ ¢ £ ¤ ¥ ¦ σ Therefore, & = ¡ ¢ £ ¤ ¥ ¦ n i n i u 1 2 2 ~ χ σ Using this concept we can write, & = ¡ ¢ £ ¤ ¥ ¦ n i n i u 1 2 2 2 ~ ˆ χ σ Note that, we loose 2 degrees of freedom because of estimation of ) , ( 1 b b , in i i i X b b Y u 1 ˆ = . Consider the random variable & & = = = = ¡ ¢ £ ¤ ¥ ¦ n i n i i i n u u 1 2 2 2 1 2 2 ˆ ) 2 ( ˆ ˆ σ σ σ σ Which we can express as 2 2 2 2 ~ ˆ ) 2 ( n n χ σ σ 2 2. Testing Intercept: Now let us put together all the tools to discuss the hypothesis testing problem 1 1 : β β = H , where 1 β is known....
View
Full
Document
This document was uploaded on 10/29/2008.
 Fall '08

Click to edit the document details