Chapter04_06

# Chapter04_06 - Model Adequacy Checking(Chap 4 6 STAT 563...

• Notes
• 101

This preview shows pages 1–15. Sign up to view the full content.

Model Adequacy Checking (Chap 4 & 6) STAT 563 Spring 2007

This preview has intentionally blurred sections. Sign up to view the full version.

Model Assumptions Recall the five assumptions Response and the predictors are approximately linear related The error term has zero mean The error term has constant variance σ 2 The errors are uncorrelated The errors are normally distributed Assumptions (4) and (5) together imply that the errors are independent random variables Assumption (5) is required for hypothesis testing and interval estimation
Use of Residuals in checking model assumptions

This preview has intentionally blurred sections. Sign up to view the full version.

Hat Matrix Recall for the model Y=X β + ε , we estimate β via least squares as With the fitted values computed as Is an n x n matrix called hat matrix . (it puts a hat on Y) Y X X X ' ) ' ( ˆ 1 - = β ' ) ' ( ' ) ' ( ˆ ˆ 1 1 X X X X H where Y H Y X X X X X Y - - = = = = β

This preview has intentionally blurred sections. Sign up to view the full version.

Hat Matrix H, the “hat” matrix, is a projection matrix of Y onto the (p+1)-dimensional space defined by the columns of X. Y H Y = ˆ Y X 1 X 2
Note All symmetric idempotent matrices are projection matrices Also note that so that h ij measures the influence of Y j on ) ) ( ( 0 ' ) ( ' ˆ )' ˆ ( ˆ ˆ 2 Φ = - = - = - = Φ = - = - - H H H H H H I because Y Y Y H H I Y Y Y Y Since Y Y Y H is idempotent j i j ij i ii n in i i i Y h Y h Y h Y h Y h Y + = + + + = .... ˆ 2 2 1 1 i Y ˆ

This preview has intentionally blurred sections. Sign up to view the full version.

Note The h ii values are often called “leverage values” . The leverage values are only a function of the predictor variables so the indicate when a set of predictor variables will induce a large effect on the corresponding predicted value of the response Rule of Thumb If h ii > 2(p+1)/n, the i th response has an undue influence on the i th predicted response (2 times average, see below) Rationale: n p n H tr h ii / ) 1 ( / ) ( + = =
Hat Matrix Properties j i h a from h b h h h h a H H ij n j ij ii ii n j ij ii - = = = = , 1 1 )) ( ( 1 0 ) ( 1 0 ) ( 1 2 2 1 2 2

This preview has intentionally blurred sections. Sign up to view the full version.

Hat Matrix For the centered model, we can define the centered Hat matrix as Where the centered model is written as ' 1 ' ) ( C C C C C X X X X H - = ε β α ε β β β ε β β + + = + - + + = + + = = C C i j ij j j j i ij k j j i X Y x x x x y 1 ) ( 0 1 0
Centered Model Write the prediction as From above, note that . ˆ ' 1 ) ( 1 ˆ 1 ˆ ˆ ' 1 ' y and s of matrix a is J where Y X X X X J n X Y C C C C C C = + = + = - α β α C C H J n H Y H J n Y + = + = 1 1 ˆ

This preview has intentionally blurred sections. Sign up to view the full version.

Centered Model If we let h ik and h cik be the elements of these matrices, we see that We can also see that The diagonal elements of H and H C are bounded as cik ik h n h + = 1 k n k cik k n k ik i y h y y h y = = + = = 1 1 ˆ 1 0 1 1 cii ii h h n
Hat Matrix Hat matrix is unaffected by the non-singular transformation on the predictors Let T be non-singular (ie, inverse exists), and let Z=XT H X X X X X T T X X XTT X T XT X T XT Z Z Z Z H Z = = = = = - - - - - - ' ) ' ( ' ' ) ' ( ) ' ( ' ' ) ' ' ( ' ) ' ( 1 1 1 1 1 1

This preview has intentionally blurred sections. Sign up to view the full version.

Hat Matrix We are interested in two transformations: Where P is the matrix of eigenvectors of the correlation matrix Note that X S =X C T S Hence the first transformation scales the centered predictor but does not alter the Hat matrix.
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern