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: Inference Under the Normal Theory GaussMarkov Linear Model I Remember our dairy cow study (2 treatments, 3 reps per trt) and our questions (Introduction, slide 3) I We’ve answered all questions except the last one: 3) When does t = [(¯ y 1 ¯ y 2 ) ( μ 1 μ 2 )] / p s 2 * 2 / n follow a T distribution? I This underlies all statistical inference about ¯ y 1 ¯ y 2 I Math Stat tells us the random variable T defined as T = Z p V /ν has the distribution known as the T distribution when: I Z has a standard normal distribution, Z ∼ N ( , 1 ) I V has a Chisquare distribution with ν d.f., V ∼ χ 2 ν I and Z and V are independent Copyright c 2011 Dept. of Statistics (Iowa State University) Statistics 511 1 / 42 Inference (cont.) I To relate our tstatistic in Q3 to the Math Stat requires answering four additional questions I 1) What is the distribution of ¯ y 1 ¯ y 2 or C ˆ β ? I 2) What is the distribution of s 2 ? I 3) If [(¯ y 1 ¯ y 2 ) ( μ 1 μ 2 )] / p s 2 * 2 / n follows a t distribution, which t distribution? I.e., how many d.f.? I 4) Are ¯ y 1 ¯ y 2 and s 2 independent? Copyright c 2011 Dept. of Statistics (Iowa State University) Statistics 511 2 / 42 Inference (cont.) The next series of slides will show that: Under the normal GM model, y = X β + , where ∼ N ( ,σ 2 I ) and C β is estimable: I the distribution of C ˆ β , the OLS estimator of C β is C ˆ β ∼ N ( C β , σ 2 C ( X X ) C ) . I s 2 has a χ 2 distribution I k = tr ( I P x ) is the d.f. associated with the distribution of s 2 I C ˆ β and s 2 are independent I Hence, (¯ y 1 ¯ y 2 ) / p s 2 * 2 / n follows a t distribution with k d.f. I Deriving these answers requires an extended discussion on the multivariate normal distribution Copyright c 2011 Dept. of Statistics (Iowa State University) Statistics 511 3 / 42 Basic Facts about Multivariate Normal Distributions I Suppose Z 1 ,..., Z n i . i . d . ∼ N ( , 1 ) and Z = [ Z 1 ,..., Z n ] . I The vector Z has the standard multivariate dn: Z ∼ N ( , I ) . I What if we shift and scale Z , i.e. multiply Z i by a i and add μ i , i.e. W i = μ i + a i Z i I As a matrix operation, W = μ + A Z . μ is a k × 1 vector of constants; A is a k × n matrix of constants. I W = μ + A Z has a multivariate normal distribution with mean μ and variance Σ = A A . I Our notation for the multivariate normal d’n will be W ∼ N ( μ , Σ) . I When Σ is nonsingular, the density of W is 1 ( 2 π ) K / 2  Σ  1 / 2 exp 1 / 2 ( W μ ) Σ 1 ( W μ ) Copyright c 2011 Dept. of Statistics (Iowa State University) Statistics 511 4 / 42 What if we apply a linear transformation, a i + b i W i , i.e. a + B W ? I If a is an ‘ × 1 vector of constants and B is a ‘ × k matrix of constants, then a + B W ∼ N ( a + B μ , B Σ B ) ....
View
Full
Document
This note was uploaded on 02/11/2012 for the course STAT 511 taught by Professor Staff during the Spring '08 term at Iowa State.
 Spring '08
 Staff

Click to edit the document details