Optimal Constants in the Rosenthal Inequality for Random Variables with Zero Odd Moments.

Similar documents
Complete Convergence and Some Maximal Inequalities for Weighted Sums of Random Variables

Abstract. 1. Introduction

PART ONE. Solutions to Exercises

Coding Theorems on New Fuzzy Information Theory of Order α and Type β

QT codes. Some good (optimal or suboptimal) linear codes over F. are obtained from these types of one generator (1 u)-

9 U-STATISTICS. Eh =(m!) 1 Eh(X (1),..., X (m ) ) i.i.d

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

ρ < 1 be five real numbers. The

Chapter 4 Multiple Random Variables

STK4011 and STK9011 Autumn 2016

Almost Sure Convergence of Pair-wise NQD Random Sequence

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Econometric Methods. Review of Estimation

Complete Convergence for Weighted Sums of Arrays of Rowwise Asymptotically Almost Negative Associated Random Variables

Internet Electronic Journal of Molecular Design

arxiv:math/ v1 [math.gm] 8 Dec 2005

Marcinkiewicz strong laws for linear statistics of ρ -mixing sequences of random variables

Chapter 5 Properties of a Random Sample

Non-uniform Turán-type problems

Solutions for HW4. x k n+1. k! n(n + 1) (n + k 1) =.

X ε ) = 0, or equivalently, lim

PROJECTION PROBLEM FOR REGULAR POLYGONS

X X X E[ ] E X E X. is the ()m n where the ( i,)th. j element is the mean of the ( i,)th., then

Inverse Estimates for Lupas-Baskakov Operators

STATISTICAL PROPERTIES OF LEAST SQUARES ESTIMATORS. x, where. = y - ˆ " 1

Rational Equiangular Polygons

Lecture 3 Probability review (cont d)

02/15/04 INTERESTING FINITE AND INFINITE PRODUCTS FROM SIMPLE ALGEBRAIC IDENTITIES

Entropy ISSN by MDPI

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions

Multivariate Transformation of Variables and Maximum Likelihood Estimation

Lecture 02: Bounding tail distributions of a random variable

A Note on the Almost Sure Central Limit Theorem in the Joint Version for the Maxima and Partial Sums of Certain Stationary Gaussian Sequences *

INDEX BOUNDS FOR VALUE SETS OF POLYNOMIALS OVER FINITE FIELDS

Chapter 8: Statistical Analysis of Simulated Data

Some identities involving the partial sum of q-binomial coefficients

. The set of these sums. be a partition of [ ab, ]. Consider the sum f( x) f( x 1)

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

SUB-BERNOULLI FUNCTIONS, MOMENT INEQUALITIES AND STRONG LAWS FOR NONNEGATIVE AND SYMMETRIZED U-STATISTICS 1. By Cun-Hui Zhang Rutgers University

A Remark on the Uniform Convergence of Some Sequences of Functions

TESTS BASED ON MAXIMUM LIKELIHOOD

D KL (P Q) := p i ln p i q i

LECTURE - 4 SIMPLE RANDOM SAMPLING DR. SHALABH DEPARTMENT OF MATHEMATICS AND STATISTICS INDIAN INSTITUTE OF TECHNOLOGY KANPUR

The Primitive Idempotents in

CHAPTER VI Statistical Analysis of Experimental Data

( ) = ( ) ( ) Chapter 13 Asymptotic Theory and Stochastic Regressors. Stochastic regressors model

Bounds for the Connective Eccentric Index

STRONG CONSISTENCY OF LEAST SQUARES ESTIMATE IN MULTIPLE REGRESSION WHEN THE ERROR VARIANCE IS INFINITE

A tighter lower bound on the circuit size of the hardest Boolean functions

On the Exchange Property for the Mehler-Fock Transform

Maximum Likelihood Estimation

A stopping criterion for Richardson s extrapolation scheme. under finite digit arithmetic.

Introduction to Probability

Continuous Distributions

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution

Part 4b Asymptotic Results for MRR2 using PRESS. Recall that the PRESS statistic is a special type of cross validation procedure (see Allen (1971))

Chapter 4 Multiple Random Variables

Lecture 4 Sep 9, 2015

Special Instructions / Useful Data

Bayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information

Class 13,14 June 17, 19, 2015

Exchangeable Sequences, Laws of Large Numbers, and the Mortgage Crisis.

Research Article A New Iterative Method for Common Fixed Points of a Finite Family of Nonexpansive Mappings

Ordinary Least Squares Regression. Simple Regression. Algebra and Assumptions.

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Extend the Borel-Cantelli Lemma to Sequences of. Non-Independent Random Variables

Strong Convergence of Weighted Averaged Approximants of Asymptotically Nonexpansive Mappings in Banach Spaces without Uniform Convexity

Finsler Geometry & Cosmological constants

Research Article Some Strong Limit Theorems for Weighted Product Sums of ρ-mixing Sequences of Random Variables

hp calculators HP 30S Statistics Averages and Standard Deviations Average and Standard Deviation Practice Finding Averages and Standard Deviations

5 Short Proofs of Simplified Stirling s Approximation

ON THE LAWS OF LARGE NUMBERS FOR DEPENDENT RANDOM VARIABLES

THE ROYAL STATISTICAL SOCIETY 2016 EXAMINATIONS SOLUTIONS HIGHER CERTIFICATE MODULE 5

Applied Mathematics and Computation

The Mathematical Appendix

ON THE ELEMENTARY SYMMETRIC FUNCTIONS OF A SUM OF MATRICES

Parameter, Statistic and Random Samples

Chapter 4 (Part 1): Non-Parametric Classification (Sections ) Pattern Classification 4.3) Announcements

Study of Correlation using Bayes Approach under bivariate Distributions

New Power Series Inequalities and Applications

Comparison of Parameters of Lognormal Distribution Based On the Classical and Posterior Estimates

A Family of Generalized Stirling Numbers of the First Kind

J P S S. A comprehensive journal of probability and statistics for theorists, methodologists, practitioners, teachers, and others

Chapter 10 Two Stage Sampling (Subsampling)

#A27 INTEGERS 13 (2013) SOME WEIGHTED SUMS OF PRODUCTS OF LUCAS SEQUENCES

Introduction to local (nonparametric) density estimation. methods

4 Inner Product Spaces

A Martingale Proof of Dobrushin s Theorem for Non-Homogeneous Markov Chains 1

LINEAR RECURRENT SEQUENCES AND POWERS OF A SQUARE MATRIX

A new Family of Distributions Using the pdf of the. rth Order Statistic from Independent Non- Identically Distributed Random Variables

Packing of graphs with small product of sizes

Point Estimation: definition of estimators

3. Basic Concepts: Consequences and Properties

ON THE LOGARITHMIC INTEGRAL

Functions of Random Variables

Q-analogue of a Linear Transformation Preserving Log-concavity

CS286.2 Lecture 4: Dinur s Proof of the PCP Theorem

Lecture Notes Types of economic variables

BAYESIAN ESTIMATOR OF A CHANGE POINT IN THE HAZARD FUNCTION

Investigating Cellular Automata

Transcription:

Optma Costats the Rosetha Iequaty for Radom Varabes wth Zero Odd Momets. The Harvard commuty has made ths artce opey avaabe. Pease share how ths access beefts you. Your story matters Ctato Ibragmov, Rustam ad Marat Ibragmov. 008. Optma costats the Rosetha equaty for radom varabes wth zero odd momets. Statstcs ad Probabty Letters 78(): 86-89. Pubshed Verso http://dx.do.org/0.06/j.sp.007.05.08 Ctabe http://rs.harvard.edu/ur-3:hul.istrepos:6446 Terms of Use Ths artce was dowoaded from Harvard Uversty s DASH repostory, ad s made avaabe uder the terms ad codtos appcabe to Other Posted Matera, as set forth at http:// rs.harvard.edu/ur-3:hul.istrepos:dash.curret.terms-ofuse#laa

OPTIMAL CONSTANTS IN THE ROSENTHAL INEQUALITY FOR RANDOM VARIABLES WITH ZERO ODD MOMENTS Marat Ibragmov Departmet of Probabty Theory, Tashet State Ecoomcs Uversty Rustam Ibragmov Departmet of Ecoomcs, Harvard Uversty Abstract. We obta estmates for the best costat the Rosetha equaty m E ξ C m Eξ E ( )max, ξ = = = for depedet radom varabes ξ,...,ξ m wth zero frst odd momets,. The estmates are sharp the extrema cases = ad =m, that s, the cases of radom varabes wth zero mea ad radom varabes wth m zero frst odd momets. Rosetha (970) proved the foowg equaty: E = t t / t ξ C( t )max E ξ, Eξ () = = The authors are gratefu to a aoymous referee for hepfu commets ad suggestos. Correspodg author. Departmet of Ecoomcs, Harvard Uversty, Lttauer Ceter, 805 Cambrdge St., Cambrdge MA 038. Phoe: (67) 496-4795. Fax: (67) 495-7730. Ema: rbragm@fas.harvard.edu

for a postve tegers ad a depedet radom varabes (r.v. s) ξ,...,ξ wth Eξ = 0, t Eξ <, =,...,, t >, where C( t ) s a costat depedg oy o t. A umber of papers have focused o refemets ad extesos of equaty () ad reated probems (see Prohorov, 96; Sazoov, 974; Nagaev ad Pes, 977; Pes, 980, 994; Pes ad Utev, 984; Johso, Schechtma ad Z, 985; Utev, 985; Htczeo, 990, 994; Bestseaya ad Utev, 99, Ibragmov ad Sharahmetov, 995, 997, 00a, b; Fge, Htczeo, Johso, Schechtma ad Z, 997; Ibragmov, 997; de a Peña ad Gé, 999; ad de a Peña, Ibragmov ad Sharahmetov, 003). Fge et a. (997) ad Ibragmov ad Sharahmetov (995, 997) derved the foowg expressos for the best costat C * sym( t ) equaty () for symmetrc r.v. s: a Γ( a) = x e 0 x dx t / t + Γ * = + C ( t ) sym, <t<4, π * t C sym ( t ) Eθ θ =, t 4, where ad θ, θ are depedet Posso r.v. s wth parameter 0.5. The proof of the expressos for C * sym( t ) Ibragmov ad Sharahmetov (995, 997) rees o the wor by Utev (985), who obtaed, amog other resuts, sharp upper ad ower bouds o E = t ξ, t 4, where ξ,...,ξ are depedet symmetrc r.v. s wth fte tth momet, terms of E ξ = t ad = Eξ t /. Bestseaya ad Utev (99) derved a smar upper boud o eve momets of sums depedet mea-zero r.v. s ξ,...,ξ, from whch the best costat geera Rosetha s equaty () the case t= ca be deduced. Usg a dfferet proof techque, the expresso for the best costat geera equaty () for eve momets t= of sums of

3 mea-zero r.v. s was depedety obtaed Ibragmov ad Sharahmetov (00a). Ibragmov ad Sharahmetov (00b) obtaed the best costat the aaogue of equaty () for oegatve r.v. s. The resuts Ibragmov ad Sharahmetov (995, 997, 00a, b) were aso preseted Ibragmov (997). de a Peña et a. derved sharp aaogues of the Burhoder Rosetha equates ad reated estmates for the expectatos of fuctos of sums of depedet oegatve r.v. s ad codtoay symmetrc martgae dffereces wth bouded codtoa momets as we as for sums of mutear forms. The preset paper deas wth estmatg the best costats the Rosetha s equaty for r.v. s wth zero frst odd momets. Namey, et C * ( t ) deote the best costats equaty () for a postve tegers ad a depedet r.v.'s s ξ,...,ξ wth Eξ = 0, s =,,...,. The the foowg theorem hods. Theorem. If t=, m N, the j r j * ( m! ) C ( m ) ( )! j j = r = =!, () where the er sum s tae over a atura m >m >...>m r ad j,...,j r satsfyg the codtos m j +...+m r j r =, j +...+j r =j, m s, =,,..., r, s =,,...,.

4 Remar. The vaue j r j ( m! ) ( m )! equaty () has a smpe j j = r = =! combatora sese (e.g., Sachov, 996): t equas the umber of parttos of a set cosstg of eemets to parts the umber of eemets whch s ot equa to s, s =,,...,. Remar. As foows from the resuts Pes ad Utev (984), Bestseaya ad Utev (99) ad Ibragmov ad Sharahmetov (997, 00a), bouds () are sharp for = ad =m; addto, whe =m, the rght-had sde of () equas to the best costat C * sym( ) the Rosetha s equaty for symmetrc r.v. s. It s aso terestg to ote that, the case =0, the expresso o the rght-had sde of (), wth the er sum tae over a atura m>m >...>m r ad j,...,j r satsfyg the codtos m j +...+m r j r =, j +...+j r =j, equas to the best costat the aaogue of equaty () for oegatve r.v. s (see Ibragmov ad Sharahmetov, 00b). Smar to Remar, the atter expresso equas to the tota umber of parttos of a set cosstg of eemets (the -th Be umber). Let us formuate some auxary resuts eeded for the proof of Theorem. The foowg emma foows from Coroary Utev (985) ad the formua represetg momets by sem varats. Lemma. Let s ξ,..., ξ be depedet r.v. s wth Eξ = 0, s =,,...,. Set A = Eξ, = E = /, =,,...,, B = A. The foowg equaty hods:, r j j Am, ( m! ) ξ ( )!, (5) r= 0 = j!

B 5 where the er sum s tae over a atura m >m >...>m r ad j,...,j r, satsfyg the codtos m j +...+m r j r =, j +...+j r =j, m s, =,,..., r, s =,,...,. Let A, B, D>0. Deote M ( m, A, B) = sup ES, ξ m, where sup s tae over postve tegers ad a depedet r.v. s s ξ,...,ξ wth Eξ = 0, s =,,..., ad fxed A, = A, BB=B; M ( m, A, B ) = sup ES, ξ m, where sup s tae over postve tegers s ad a depedet r.v. s ξ,...,ξ wth Eξ = 0, s =,,...,, for whch A, A, B B; M ( m, D) = sup ES m, where sup s tae over postve tegers ad a depedet r.v. s, ξ s ξ,...,ξ wth Eξ = 0, s =,,...,, ad fxed max( A,, B ) = D. The foowg emma s we-ow (see, e.g., Pes ad Utev, 984). Lemma. For <s< m s ( m s) ( ) A A B s,, /( ). (6) Reato (5) ad Lemma mpy the foowg Lemma 3. For A, B>0

6 m M j r j (!) m A m B m j ( A m j m B m j /( m ) ( ) (,, ) ( )! ) j r!, = = = =,, where the er sum s tae over a atura m >m >...>m r ad j,...,j r, satsfyg the codtos m j +...+m r j r =, j +...+j r =j, m s, =,,..., r, s =,,...,. Proof of Theorem. From Lemma 3 ad the evdet equaty / m M ( m, D) M ( m, D, D ) t foows that M ( m, D) ( )! j j = r = r = ( m!) j j! D, where the er sum s tae over a atura m >m >...>m r ad j,..., j r, satsfyg the codtos m j +...+m r j r =, j +...+j r =j, m s, =,,..., r, s =,,...,. Sce * M ( m, D) C ( ) = sup, (7) D ths mpes (). D> 0

7 REFERENCES Bestseaya, E. V. ad Utev, S. A. (99). A exact upper boud for the eve momet of sums of depedet radom varabes. Sbera Math. J. 3, 39-4. de a Peña, V. H. ad Gé, E. (999). Decoupg. From depedece to depedece. Radomy stopped processes. U-statstcs ad processes. Martgaes ad beyod. Probabty ad ts Appcatos (New Yor). Sprger-Verag, New Yor. de a Peña, V. H., Ibragmov, R. ad Sharahmetov, S. (003). O extrema dstrbutos ad sharp Lp -bouds for sums of mutear forms. A. Probab. 3, 630-675. Fge, T., Htczeo, P., Johso, W. B., Schechtma, G. ad Z, J. (997). Extrema propertes of Rademacher fuctos wth appcatos to the Khtche ad Rosetha equates. Tras. Amer. Math. Soc., 349, 997-07. Htczeo, P. (990). Best costats martgae verso of Rosetha s equaty. A. Probab. 8, 656-668. Htczeo, P. (994). O a domato of sums of radom varabes by sums of codtoay depedet oes. A. Probab., 453-468. Ibragmov, R. (997). Estmates for the momets of symmetrc statstcs. Ph.D. Dssertato. Isttute of Mathematcs of Uzbe Academy of Sceces, Tashet, 7 pp. ( Russa). Ibragmov, R. ad Sharahmetov, Sh. (995). O the best costat Rosetha's equaty. I: Theses of reports of the coferece o probabty theory ad mathematca statstcs dedcated to the 75th aversary of Academca S. Kh. Srajdov (Fergaa, Uzbesta). Tashet, 43-44 ( Russa).

8 Ibragmov, R. ad Sharahmetov, Sh. (997). O a exact costat for the Rosetha equaty. Teor. Veroyatost. Prme. 4, 34-350 (trasato Theory Probab. App. 4 (997), 94-30 (998)). Ibragmov, R. ad Sharahmetov, S. (00a). The exact costat the Rosetha equaty for radom varabes wth mea zero. Theory Probab. App. 46, 7-3. Ibragmov, R. ad Sharahmetov, S. (00b). The best costat the Rosetha equaty for oegatve radom varabes. Statst. Probab. Lett. 55, 367-376. Johso, W. B., Schechtma, G. ad Z, J. (985). Best costats momet equates for ear combatos of depedet ad exchageabe radom varabes. A. Probab. 3, 34-53. Nagaev, S. V. ad Pes, I. F. (977). Some equates for the dstrbutos of sums of depedet radom varabes. Theory of Probab. App., 48-56. Pes, I. F. (980). Estmates for momets of fte-dmesoa martgaes. Math. Notes 7, 459-46. Pes, I. (994). Optmum bouds for the dstrbutos of martgaes Baach spaces. A. Probab., 679--706. Pes, I. F. ad Utev, S. A. (984). Estmates of momets of sums of depedet radom varabes. Theory Probab. App. 9, 574-577. Prohorov, Yu. V. (96). Extrema probems mt theorems. I Proc. VI A-Uo Coferece o Probabty Theory ad Mathematca Statstcs, Vus, 77-84. Rosetha, H. P. (970). O the subspaces of L p (p>) spaed by sequeces of depedet radom varabes. Israe J. Math. 8, 73-303. Sachov, V. N. (966). Combatora methods dscrete mathematcs. Ecycopeda of Mathematcs ad ts Appcatos, 55. Cambrdge Uversty Press, Cambrdge, 306 pp.

9 Sazoov, V. V. (974). O the estmato of momets of sums of depedet radom varabes. Theory Probab. App. 9, 37-374. Utev, S. A. (985). Extrema probems momet equates. I Proc. Mathematca Isttute of the Sbera Brach of the USSR Academy of Sceces, 5, 56-75 ( Russa).