Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 The Bootstrap’s Finite Sample Distribution An Analytical Approach Lawrence C. Marsh Department of Economics and Econometrics University of Notre Dame.

Similar presentations


Presentation on theme: "1 The Bootstrap’s Finite Sample Distribution An Analytical Approach Lawrence C. Marsh Department of Economics and Econometrics University of Notre Dame."— Presentation transcript:

1 1 The Bootstrap’s Finite Sample Distribution An Analytical Approach Lawrence C. Marsh Department of Economics and Econometrics University of Notre Dame Midwest Econometrics Group (MEG) October 15 – 16, 2004 Northwestern University

2 2 This is the first of three papers: (1.) Bootstrap’s Finite Sample Distribution ( today !!! ) (2.) Bootstrapped Asymptotically Pivotal Statistics (3.) Bootstrap Hypothesis Testing and Confidence Intervals

3 3 traditional approach in econometrics Analytical solution Bootstrap’s Finite Sample Distribution Empirical process  approach used in this paper Analytical problem Analogy principle (Manski) GMM (Hansen) Empirical process 

4 4 Bootstrap sample of size m: Start with a sample of size n:{X i : i = 1,…,n} {X j * : j = 1,…,m} m n Define M i as the frequency of drawing each X i. bootstrap procedure

5 5...

6 6 for i  k

7 7 = Applied Econometrician: The bootstrap treats the original sample as if it were the population and induces multinomial distributed randomness.

8 8 = Econometric theorist: what does this buy you? Find out under joint distribution of bootstrap-induced randomness and randomness implied by the original sample data:

9 9 = Econometric theorist: Applied Econometrician: For example,

10 10 The Wild Bootstrap Multiply each boostrapped value by plus one or minus one each with a probability of one-half (Rademacher Distribution). Use binomial distribution to impose Rademacher distribution: W i = number of positive ones out of M i which, in turn, is the number of X i ’s drawn in m multinomial draws.

11 11 The Wild Bootstrap = Econometric Theorist: Applied Econometrician: under zero mean assumption

12 12...

13 13 almost surely, where is matrix of second partial derivatives of g. where X is a p x 1 vector. nonlinear function of . Horowitz (2001) approximates the bias of for a smooth nonlinear function gas an estimator of

14 14

15 15 Horowitz (2001) uses bootstrap simulations to approximate the first term on the right hand side. Exact finite sample solution: = +

16 16 Definition: Any bootstrap statistic,, that is a function of the elements of the set {f(X j * ): j = 1,…,m } and satisfies the separability condition where g(M i ) and h( f(X i )) are independent functions and where the expected value E M [g(M i )] exists, is a “directly analyzable” bootstrap statistic. Separability Condition

17 17 X is an n x 1 vector of original sample values. X * is an m x 1 vector of bootstrapped sample values. X * = HX where the rows of H are all zeros except for a one in the position corresponding to the element of X that was randomly drawn. E H [H] = (1/n) 1 m 1 n ’ where 1 m and 1 n are column vectors of ones.  m * = g(X * ) = g(HX ) Taylor series expansion  m * = g(X o * ) + [G 1 (X o * )]’(X *  X o * ) + (1/2) (X *  X o * )’[G 2 (X o * )](X *  X o * ) + R * Setup for empirical process: X o * = H o X

18 18  m * = g(X * ) = g(HX ) Taylor series expansion  m * = g(( 1/n )1 m 1 n ’X ) + [G 1 (( 1/n )1 m 1 n ’X )]’(H  ( 1/n )1 m 1 n ’) X + ( 1/2 )X ‘(H  ( 1/n )1 m 1 n ’)’[G 2 (( 1/n )1 m 1 n ’X )](H  ( 1/n )1 m 1 n ’) X + R * Taylor series: Now ready to determine exact finite moments, et cetera. X * = HX where the rows of H are all zeros except for a one in the position corresponding to the element of X that was randomly drawn. Setup for analytical solution: X o * = H o X H o = E H [H] = (1/n) 1 m 1 n ’

19 19 e = ( I n – X (X’X) -1 X’)  {,,..., } E H [H] = (1/n) 1 n 1 n ’ e * = H e A = ( I n – (1/n)1 n 1 n ’ ) } No restrictions on covariance matrix for errors.

20 20 Applied Econometrician:. = A = I n or where A1 n 1 n ’ = 01 n 1 n ’A = 0 and A = ( I n – (1/n)1 n 1 n ’ ) so

21 21 Econometric theorist: + where No restrictions on

22 22 This is the first of three papers: (1.) Bootstrap’s Finite Sample Distribution ( today !!! ) (2.) Bootstrapped Asymptotically Pivotal Statistics (3.) Bootstrap Hypothesis Testing and Confidence Intervals Thank you ! basically done. almost done.


Download ppt "1 The Bootstrap’s Finite Sample Distribution An Analytical Approach Lawrence C. Marsh Department of Economics and Econometrics University of Notre Dame."

Similar presentations


Ads by Google