Journal of Inequalities in Pure and Applied Mathematics

Similar documents
Banach Journal of Mathematical Analysis ISSN: (electronic)

Asymptotically Lacunary Statistical Equivalent Sequence Spaces Defined by Ideal Convergence and an Orlicz Function

SOME SOLVABILITY THEOREMS FOR NONLINEAR EQUATIONS

ON INDEPENDENT SETS IN PURELY ATOMIC PROBABILITY SPACES WITH GEOMETRIC DISTRIBUTION. 1. Introduction. 1 r r. r k for every set E A, E \ {0},

FRACTIONAL HERMITE-HADAMARD TYPE INEQUALITIES FOR FUNCTIONS WHOSE SECOND DERIVATIVE ARE

JENSEN S INEQUALITY FOR DISTRIBUTIONS POSSESSING HIGHER MOMENTS, WITH APPLICATION TO SHARP BOUNDS FOR LAPLACE-STIELTJES TRANSFORMS

SOME GENERAL NUMERICAL RADIUS INEQUALITIES FOR THE OFF-DIAGONAL PARTS OF 2 2 OPERATOR MATRICES

Do Managers Do Good With Other People s Money? Online Appendix

HE DI ELMONSER. 1. Introduction In 1964 H. Mink and L. Sathre [15] proved the following inequality. n, n N. ((n + 1)!) n+1

Functions Defined on Fuzzy Real Numbers According to Zadeh s Extension

STUDY OF SOLUTIONS OF LOGARITHMIC ORDER TO HIGHER ORDER LINEAR DIFFERENTIAL-DIFFERENCE EQUATIONS WITH COEFFICIENTS HAVING THE SAME LOGARITHMIC ORDER

KOEBE DOMAINS FOR THE CLASSES OF FUNCTIONS WITH RANGES INCLUDED IN GIVEN SETS

Research Article On Alzer and Qiu s Conjecture for Complete Elliptic Integral and Inverse Hyperbolic Tangent Function

A generalization of the Bernstein polynomials

A Short Combinatorial Proof of Derangement Identity arxiv: v1 [math.co] 13 Nov Introduction

q i i=1 p i ln p i Another measure, which proves a useful benchmark in our analysis, is the chi squared divergence of p, q, which is defined by

ON LACUNARY INVARIANT SEQUENCE SPACES DEFINED BY A SEQUENCE OF MODULUS FUNCTIONS

arxiv: v1 [math.co] 4 May 2017

On the ratio of maximum and minimum degree in maximal intersecting families

On the Quasi-inverse of a Non-square Matrix: An Infinite Solution

Banach Journal of Mathematical Analysis ISSN: (electronic)

Results on the Commutative Neutrix Convolution Product Involving the Logarithmic Integral li(

arxiv: v1 [math.ca] 31 Aug 2009

CENTRAL INDEX BASED SOME COMPARATIVE GROWTH ANALYSIS OF COMPOSITE ENTIRE FUNCTIONS FROM THE VIEW POINT OF L -ORDER. Tanmay Biswas

arxiv: v1 [math.nt] 12 May 2017

Solution to HW 3, Ma 1a Fall 2016

9.1 The multiplicative group of a finite field. Theorem 9.1. The multiplicative group F of a finite field is cyclic.

Lacunary I-Convergent Sequences

arxiv: v1 [math.co] 6 Mar 2008

On the ratio of maximum and minimum degree in maximal intersecting families

On the Poisson Approximation to the Negative Hypergeometric Distribution

A THREE CRITICAL POINTS THEOREM AND ITS APPLICATIONS TO THE ORDINARY DIRICHLET PROBLEM

A NOTE ON VERY WEAK SOLUTIONS FOR A CLASS OF NONLINEAR ELLIPTIC EQUATIONS

Measure Estimates of Nodal Sets of Polyharmonic Functions

Mean Curvature and Shape Operator of Slant Immersions in a Sasakian Space Form

Math 301: The Erdős-Stone-Simonovitz Theorem and Extremal Numbers for Bipartite Graphs

ESSENTIAL NORM OF AN INTEGRAL-TYPE OPERATOR ON THE UNIT BALL. Juntao Du and Xiangling Zhu

Numerical approximation to ζ(2n+1)

Semicanonical basis generators of the cluster algebra of type A (1)

JANOWSKI STARLIKE LOG-HARMONIC UNIVALENT FUNCTIONS

A Multivariate Normal Law for Turing s Formulae

COLLAPSING WALLS THEOREM

RADIAL POSITIVE SOLUTIONS FOR A NONPOSITONE PROBLEM IN AN ANNULUS

Chromatic number and spectral radius

ON THE INVERSE SIGNED TOTAL DOMINATION NUMBER IN GRAPHS. D.A. Mojdeh and B. Samadi

New problems in universal algebraic geometry illustrated by boolean equations

Some Ideal Convergent Sequence Spaces Defined by a Sequence of Modulus Functions Over n-normed Spaces

ONE-POINT CODES USING PLACES OF HIGHER DEGREE

Fractional Zero Forcing via Three-color Forcing Games

Relating Branching Program Size and. Formula Size over the Full Binary Basis. FB Informatik, LS II, Univ. Dortmund, Dortmund, Germany

Using Laplace Transform to Evaluate Improper Integrals Chii-Huei Yu


Divisibility. c = bf = (ae)f = a(ef) EXAMPLE: Since 7 56 and , the Theorem above tells us that

ON SPARSELY SCHEMMEL TOTIENT NUMBERS. Colin Defant 1 Department of Mathematics, University of Florida, Gainesville, Florida

Bifurcation Analysis for the Delay Logistic Equation with Two Delays

Miskolc Mathematical Notes HU e-issn Tribonacci numbers with indices in arithmetic progression and their sums. Nurettin Irmak and Murat Alp

A NOTE ON ROTATIONS AND INTERVAL EXCHANGE TRANSFORMATIONS ON 3-INTERVALS KARMA DAJANI

Lecture 28: Convergence of Random Variables and Related Theorems

On decompositions of complete multipartite graphs into the union of two even cycles

A STABILITY RESULT FOR p-harmonic SYSTEMS WITH DISCONTINUOUS COEFFICIENTS. Bianca Stroffolini. 0. Introduction

A Bijective Approach to the Permutational Power of a Priority Queue

arxiv: v1 [math.nt] 28 Oct 2017

Central Coverage Bayes Prediction Intervals for the Generalized Pareto Distribution

Surveillance Points in High Dimensional Spaces

Math 124B February 02, 2012

Unobserved Correlation in Ascending Auctions: Example And Extensions

Integral operator defined by q-analogue of Liu-Srivastava operator

Secret Exponent Attacks on RSA-type Schemes with Moduli N = p r q

Vanishing lines in generalized Adams spectral sequences are generic

On absence of solutions of a semi-linear elliptic equation with biharmonic operator in the exterior of a ball

International Journal of Mathematical Archive-3(12), 2012, Available online through ISSN

Quasi-Randomness and the Distribution of Copies of a Fixed Graph

Multiple Criteria Secretary Problem: A New Approach

On the integration of the equations of hydrodynamics

Available online through ISSN

SPECTRAL SEQUENCES. im(er

On the global uniform asymptotic stability of time-varying dynamical systems

arxiv: v2 [math.ag] 4 Jul 2012

Journal of Number Theory

The Chromatic Villainy of Complete Multipartite Graphs

Web-based Supplementary Materials for. Controlling False Discoveries in Multidimensional Directional Decisions, with

The Archimedean Circles of Schoch and Woo

Hypothesis Test and Confidence Interval for the Negative Binomial Distribution via Coincidence: A Case for Rare Events

3.1 Random variables

Doubling property for the Laplacian and its applications (Course Chengdu 2007)

MATH 417 Homework 3 Instructor: D. Cabrera Due June 30. sin θ v x = v r cos θ v θ r. (b) Then use the Cauchy-Riemann equations in polar coordinates

Application of Parseval s Theorem on Evaluating Some Definite Integrals

On Computing Optimal (Q, r) Replenishment Policies under Quantity Discounts

Pearson s Chi-Square Test Modifications for Comparison of Unweighted and Weighted Histograms and Two Weighted Histograms

Enumerating permutation polynomials

Dalimil Peša. Integral operators on function spaces

GROWTH ESTIMATES THROUGH SCALING FOR QUASILINEAR PARTIAL DIFFERENTIAL EQUATIONS

6 PROBABILITY GENERATING FUNCTIONS

Monotonluğu Koruyan Matrisler. Monotonicity Preserving Matrices

Quadratic Harmonic Number Sums

KANTOROVICH TYPE INEQUALITIES FOR THE DIFFERENCE WITH TWO NEGATIVE PARAMETERS. Received April 13, 2010; revised August 18, 2010

NOTE. Some New Bounds for Cover-Free Families

On Polynomials Construction

Compactly Supported Radial Basis Functions

THE JEU DE TAQUIN ON THE SHIFTED RIM HOOK TABLEAUX. Jaejin Lee

Transcription:

Jounal of Inequalities in Pue and Applied Mathematics COEFFICIENT INEQUALITY FOR A FUNCTION WHOSE DERIVATIVE HAS A POSITIVE REAL PART S. ABRAMOVICH, M. KLARIČIĆ BAKULA AND S. BANIĆ Depatment of Mathematics Univesity of Haifa Haifa, 31905, Isael. EMail: abamos@math.haifa.ac.il Depatment of Mathematics Faculty of Natual Sciences, Mathematics and Education Univesity of Split Teslina 12, 21000 Split Coatia. EMail: milica@pmfst.h Faculty of Civil Engineeing and Achitectue Univesity of Split Matice hvatske 15, 21000 Split Coatia. EMail: Senka.Banic@gadst.h volume 7, issue 2, aticle 70, 2006. Received 03 Januay, 2006; accepted 20 Januay, 2006. Communicated by: C.P. Niculescu Abstact Home Page c 2000 Victoia Univesity ISSN electonic: 1443-5756 002-06

Abstact A vaiant of Jensen-Steffensen s inequality is consideed fo convex and fo supequadatic functions. Consequently, inequalities fo powe means involving not only positive weights have been established. 2000 Mathematics Subject Classification: 26D15, 26A51. Key wods: Jensen-Steffensen s inequality, Monotonicity, Supequadaticity, Powe means. 1 Intoduction......................................... 3 2 Vaiants of fo Positive Supequadatic................................ 10 3 An Altenative Poof of Theoem B..................... 19 Refeences Page 2 of 27

1. Intoduction. Let I be an inteval in R and f : I R a convex function on I. If ξ = ξ 1,..., ξ m is any m-tuple in I m and p = p 1,..., p m any nonnegative m- tuple such that m p i > 0, then the well known Jensen s inequality see fo example [7, p. 43] 1.1 f 1 P m p i ξ i 1 P m p i f ξ i holds, whee P m = m p i. If f is stictly convex, then 1.1 is stict unless ξ i = c fo all i {j : p j > 0}. It is well known that the assumption p is a nonnegative m-tuple can be elaxed at the expense of moe estictions on the m-tuple ξ. If p is a eal m-tuple such that 1.2 0 P j P m, j = 1,..., m, P m > 0, whee P j = j p i, then fo any monotonic m-tuple ξ inceasing o deceasing in I m we get ξ = 1 p i ξ i I, P m and fo any function f convex on I 1.1 still holds. Inequality 1.1 consideed unde conditions 1.2 is known as the Jensen-Steffensen s inequality [7, p. 57] fo convex functions. In his pape [5] A. McD. Mece consideed some monotonicity popeties of powe means. He poved the following theoem: Page 3 of 27

Theoem A. Suppose that 0 < a < b and a x 1 x 2 x n b hold with at least one of the x k satisfying a < x k < b. If w = w 1,..., w n is a positive n-tuple with n w i = 1 and < < s < +, then a < Q a, b; x < Q s a, b; x < b, whee fo all eal t 0, and Q t a, b; x Q 0 a, b; x ab G, a t + b t w i x t i G = n x w i i. In his next pape [6], Mece gave a vaiant of Jensen s inequality fo which Witkowski pesented in [8] a shote poof. This is stated in the following theoem: Theoem B. If f is a convex function on an inteval containing an n-tuple x = x 1,..., x n such that 0 < x 1 x 2 x n and w = w 1,..., w n is a positive n-tuple with n w i = 1, then f x 1 + x n w i x i f x 1 + f x n w i f x i. This theoem is a special case of the following theoem poved in [4] by Abamovich, Klaičić Bakula, Matić and Pečaić: 1 t Page 4 of 27

Theoem C [4, Th. 2]. Let f : I R, whee I is an inteval in R and let [a, b] I, a < b. Let x = x 1,..., x n be a monotonic n-tuple in [a, b] n and v = v 1,..., v n a eal n tuple such that v i 0, i = 1,..., n, and 0 V j V n, j = 1,..., n, V n > 0, whee V j = j v i. If f is convex on I, then 1.3 f a + b 1 v i x i f a + f b 1 v i f x i. V n V n In case f is stictly convex, the equality holds in 1.3 iff one of the following two cases occus: 1. eithe x = a o x = b, 2. thee exists l {2,..., n 1} such that x = x 1 + x n x l and 1.4 x 1 = a, x n = b o x 1 = b, x n = a, V j x j 1 x j = 0, j = 2,..., l, V j x j x j+1 = 0, j = l,..., n 1, whee V j = n i=j v i, j = 1,..., n and x = 1/V n n v ix i. In the special case whee v > 0 and f is stictly convex, the equality holds in 1.4 iff x i = a, i = 1,..., n, o x i = b, i = 1,..., n. Hee, as in the est of the pape, when we say that an n-tuple ξ is inceasing deceasing we mean that ξ 1 ξ 2 ξ n ξ 1 ξ 2 ξ n. Similaly, Page 5 of 27

when we say that a function f : I R is inceasing deceasing on I we mean that fo all u, v I we have u < v fu fv u < v fu fv. In Section 2 we efine Theoems A, B, and C. These efinements ae achieved by supequadatic functions which wee intoduced in [1] and [2]. As Jensen s inequality fo convex functions is a genealization of Hölde s inequality fo f x = x p, p 1, so the inequalities satisfied by supequadatic functions ae genealizations of the inequalities satisfied by the supequadatic functions f x = x p, p 2 see [1], [2]. Fist we quote some definitions and state a list of basic popeties of supequadatic functions. Definition 1.1. A function f : [0, R is supequadatic povided that fo all x 0 thee exists a constant Cx R such that 1.5 f y f x f y x C x y x fo all y 0. Definition 1.2. A function f : [0, R is said to be stictly supequadatic if 1.5 is stict fo all x y whee xy 0. Lemma A [2, Lemma 2.3]. Suppose that f is supequadatic. Let ξ i 0, i = 1,..., m, and let ξ = m p iξ i, whee p i 0, i = 1,..., m, and m p i = 1. Then p i f ξ i f ξ p i f ξi ξ. Lemma B [1, Lemma 2.2]. Let f be supequadatic function with Cx as in Definition 1.1. Then: Page 6 of 27

i f0 0, ii if f0 = f 0 = 0 then Cx = f x wheneve f is diffeentiable at x > 0, iii if f 0, then f is convex and f0 = f 0 = 0. In [3] the following efinement of Jensen s Steffensen s type inequality fo nonnegative supequadatic functions was poved: Theoem D [3, Th. 1]. Let f : [0, [0, be a diffeentiable and supequadatic function, let ξ be a nonnegative monotonic m-tuple in R m and p a eal m-tuple, m 3, satisfying Let ξ be defined as Then 1.6 0 P j P m, j = 1,..., m, P m > 0. p i f ξ i P m f ξ ξ = 1 P m p i ξ i. k 1 P i f ξ i+1 ξ i + P k f ξ ξ k + P k+1 f ξ k+1 ξ + P i f ξ i ξ i 1 i=k+2 Page 7 of 27

[ P i + i=k+1 P i ] f m p i ξ ξi k P i + m m m 1 P m f p i ξ ξi, m 1 P m whee P i = m j=i p j and k {1,..., m 1} satisfies i=k+1 P i ξ k ξ ξ k+1. In case f is also stictly supequadatic, inequality p i f ξ i P m f ξ m > m 1 P m f p ξi i ξ m 1 P m holds fo ξ > 0 unless one of the following two cases occus: 1. eithe ξ = ξ 1 o ξ = ξ m, 2. thee exists k {3,..., m 2} such that ξ = ξ k and { Pj ξ j ξ j+1 = 0, j = 1,..., k 1 1.7 P j ξ j ξ j 1 = 0, j = k + 1,..., m. In these two cases p i f ξ i P m f ξ = 0. Page 8 of 27

In Section 2 we efine Theoem B and Theoem C fo functions which ae supequadatic and positive. One of the efinements is deived easily fom Theoem D. We use in Section 3 the following theoem [7, p. 323] to give an altenative poof of Theoem B. Theoem E. Let I be an inteval in R, and ξ, η two deceasing m-tuples such that ξ, η I m. Let p be a eal m-tuple such that 1.8 p i ξ i p i η i fo k = 1, 2,..., m 1, and 1.9 p i ξ i = p i η i. Then fo evey continuous convex function f : I R we have 1.10 p i f ξ i p i f η i. Page 9 of 27

2. Vaiants of fo Positive Supequadatic In this section we efine in two ways Theoem C fo functions which ae supequadatic and positive. The efinement in Theoem 2.1 follows by showing that it is a special case of Theoem D fo specific p. The efinement in Theoem 2.2 follows the steps in the poof of Theoem B given by Witkowski in [8]. Theefoe the second efinement is confined only to the specific p given in Theoem B, which means that what we get is a vaiant of Jensen s inequality and not of the moe geneal Jensen-Steffensen s inequality. Theoem 2.1. Let f : [0, [0, and let [a, b] [0,. Let x = x 1,..., x n be a monotonic n tuple in [a, b] n and v = v 1,..., v n a eal n- tuple such that v i 0, i = 1,..., n, 0 V j V n, j = 1,..., n, and V n > 0, whee V j = j v i. If f is diffeentiable and supequadatic, then 2.1 f a + f b 1 v i f x i f a + b 1 v i x i V n V n n + 1 f b a 1 n V n v i a + b x i 1 n V n j=1 v jx j. n + 1 In case f is also stictly supequadatic and a > 0, inequality 2.1 is stict unless one of the following two cases occus: 1. eithe x = a o x = b, Page 10 of 27

2. thee exists l {2,..., n 1} such that x = x 1 + x n x l and x 1 = a, x n = b o x 1 = b, x n = a 2.2 V j x j 1 x j = 0, j = 2,..., l, V j x j x j+1 = 0, j = l,..., n 1, whee V j = n i=j v i, j = 1,..., n, and x = 1 n V n v ix i. In these two cases we have f a + f b 1 v i f x i f a + b 1 v i x i = 0. V n V n In the special case whee v > 0 and f is also stictly supequadatic, the equality holds in 2.1 iff x i = a, i = 1,..., n, o x i = b, i = 1,..., n. Poof. Suppose that x is an inceasing n-tuple in [a, b] n. The poof of the theoem is an immediate esult of Theoem D, by defining the n + 2-tuples ξ and p as ξ 1 = a, ξ i+1 = x i, i = 1,..., n, ξ n+2 = b p 1 = 1, p i+1 = v i /V n, i = 1,..., n, p n+2 = 1. Then we get 2.1 fom the last inequality in 1.6 and fom the fact that in ou special case we have P i + n+2 i=k+1 P i n + 1, Page 11 of 27

fo P i = i j=1 p j and P i = n+2 j=i p j, and ξ = 1 P m p i ξ i = a + b 1 V n v i x i = a + b x. The poof of the equality case and the special case whee v > 0 follows also fom Theoem D. We have P j = V j V n, j = 1,..., n, P n+1 = 0, P n+2 = 1, P 1 = 1, P 2 = 0, P j = V j 2 V n, j = 3,..., n + 2. Obviously, ξ = ξ 1 is equivalent to x = b and ξ = ξ n+2 is equivalent to x = a. Also, the existence of some k {3,..., m 2} such that ξ = ξ k and that 1.7 holds is equivalent to the existence of some l {2,..., n 1} such that x = x 1 +x n x l = a+b x l and that 2.2 holds. Theefoe, applying Theoem D we get the desied conclusions. In the case when x is deceasing we simply eplace x and v with x = x n,..., x 1 and ṽ = v n,..., v 1, espectively, and then ague in the same manne. In the special case that v > 0 also, V i > 0 and V j > 0, i = 1,..., n, and theefoe accoding to 2.2 equality holds in 2.1 only when eithe x 1 = = x n = a o x 1 = = x n = b. In the following theoem we will pove a efinement of Theoem B. Without loss of geneality we assume that n v i = 1. Page 12 of 27

Theoem 2.2. Let f : [0, [0, and let [a, b] [0,, a < b. Let x = x 1,..., x n be an n-tuple in [a, b] n and v = v 1,..., v n a eal n-tuple such that v 0 and n v i = 1. If f is supequadatic we have f a + f b v i f x i f a + b v i x i 2.3 v i f v j x j x i j=1 [ xi a + 2 v i b a f b x i + b x ] i b a f x i a 2 xi a b x i v i f v j x j x i + 2 v i f. b a j=1 If f is stictly supequadatic and v > 0 equality holds in 2.3 iff x i = a, i = 1,..., n, o x i = b, i = 1,..., n. Poof. The poof follows the technique in [8] and efines the esult to positive supequadatic functions. Fom Lemma A we know that fo any λ [0, 1] the following holds: 2.4 λf a + 1 λ f b f λa + 1 λ b λf a λa 1 λ b + 1 λ f b λa 1 λ b = λf 1 λ a b + 1 λ f λ b a = λf 1 λ b a + 1 λ f λ b a. Page 13 of 27

Also, fo any x i [a, b] thee exists a unique λ i [0, 1] such that x i = λ i a + 1 λ i b. We have 2.5 f a+f b v i f x i = f a+f b v i f λ i a + 1 λ i b. Applying 2.4 on evey x i = λ i a + 1 λ i b in 2.5 we obtain f a + f b v i f x i 2.6 = f a + f b + [ v i λi f a 1 λ i f b + λ i f 1 λ i b a + 1 λ i f λ i b a ] v i [1 λ i f a + λ i f b] + v i [λ i f 1 λ i b a + 1 λ i f λ i b a]. Applying again 2.4 on 2.6 we get 2.7 f a + f b v i f x i v i f 1 λ i a + λ i b + 2 v i [λ i f 1 λ i b a + 1 λ i f λ i b a]. Page 14 of 27

Applying again Lemma A on 2.7 we obtain 2.8 f a + f b v i f x i f v i [1 λ i a + λ i b] = f + + 2 v i f 1 λ i a + λ i b v j [1 λ j a + λ j b] j=1 v i [1 λ i f λ i b a + λ i f 1 λ i b a] a + b v i x i + v i f v j x j x i j=1 [ xi a + 2 v i b a f b x i + b x ] i b a f x i a, and this is the fist inequality in 2.3. Since f is a nonnegative supequadatic function, fom Lemma B we know that it is also convex, so fom 2.8 we have [ xi a v i b a f b x i + b x ] i b a f x i a hence, the second inequality in 2.3 is poved. 2 b xi x i a v i f, b a Page 15 of 27

Fo the case when f is stictly supequadatic and v > 0 we may deduce that inequalities 2.6 and 2.7 become equalities iff each of the λ i, i = 1,..., n, is eithe equal to 1 o equal to 0, which means that x i {a, b}, i = 1,..., n. Howeve, since we also have v j x j x i = 0, i = 1,..., n, j=1 we deduce that x i = a, i = 1,..., n, o x i = b, i = 1,..., n. This completes the poof of the theoem. Coollay 2.3. Let v = v 1,..., v n be a eal n-tuple such that v 0, n v i = 1 and let x = x 1,..., x n be an n-tuple in [a, b] n, 0 < a < b. Then fo any eal numbes and s such that s 2 we have 2.9 s Qs a, b; x 1 Q a, b; x s 1 Q a, b; x s v i v j x j x i j=1 ] x +2 v i a i b a b x i s b x + i b a x i a s s 1 Q a, b; x s v i v j x j x i j=1 Page 16 of 27

+2 v i 2 x i a b x i b a ] s, whee Q p a, b; x = a p + b p n v ix i 1 p, p R \ {0}. If s > 2 and v > 0, the equalities hold in 2.9 iff x i = a, i = 1,..., n o x i = b, i = 1,..., n. Poof. We define a function f : 0, 0, as f x = x s. It can be easily checked that fo any eal numbes and s such that s 2 the function f is supequadatic. We define a new positive n-tuple ξ in [a, b ] as ξ i = x i, i = 1,..., n. Fom Theoem 2.2 we have s a s + b s v i x s i a + b v i x i 2.10 v i v j x j x i j=1 s [ x + 2 v i a i b a b x i s s v i v j x j x i + 2 j=1 v i 2 x i a b x i b a ] b x + i b a x i a s s 0. Page 17 of 27

We have a s + b s v i x s i a + b v i x i s = Q s a, b; x s Q a, b; x s so fom 2.10 the inequalities in 2.9 follow. The equality case follows fom the equality case in Theoem 2.2, as the function f x = x s is stictly supequadatic fo s > 2. Remak 1. It is an immediate esult of Coollay 2.3 that if s > 2 and thee is at least one j {1,..., n} such that then fo this j we have s Qs a, b; x 1 > Q a, b; x v j x j a b x j > 0, 2v j Q a, b; x s 2 x j a b x j b a s > 0. Page 18 of 27

3. An Altenative Poof of Theoem B In this section we give an inteesting altenative poof of Theoem B based on Theoem E. To cay out that poof we need the following technical lemma. Lemma 3.1. Let y = y 1,..., y m be a deceasing eal m-tuple and p = p 1,..., p m a nonnegative eal m-tuple with m p i = 1. We define y = p i y i and the m-tuple y = y, y,..., y. Then the m-tuples η = y, ξ = y and p satisfy conditions 1.8 and 1.9. Poof. Note that y is a convex combination of y 1, y 2,..., y m, so we know that y m y y 1. Fom the definitions of the m-tuples ξ and η we have p i ξ i = y p i = y = p i y i = p i η i. Hence, condition 1.9 is satisfied. Futhemoe, fo k = 1, 2,..., m 1 we Page 19 of 27

have p i η i p i ξ i = = p i y i y p i y i j=1 p i p j y j p i. Since m p i = 1, we can wite p i η i p i ξ i = p j + = = = j=1 j=k+1 p j p i p i p j j=k+1 p i y i m j=k+1 m j=k+1 p j y i p i j=k+1 p j y i y j. p i y i p j y j + m j=k+1 p j y j p j y j j=1 j=k+1 p j y j p i Page 20 of 27

Since p is nonnegative and y is deceasing, we obtain p i η i p i ξ i 0, k = 1, 2,..., m 1, which means that condition 1.8 is satisfied as well. Now we can give an altenative poof of Theoem B which is mainly based on Theoem E. Poof of Theoem B. Since x = n w ix i is a convex combination of x 1, x 2,..., x n it is clea that thee is an s {1, 2,..., n 1} such that that is, x 1 x s x x s+1 x n, 3.1 x 1 x s x x s+1 x n. Adding x 1 + x n to all the inequalities in 3.1 we obtain x n x 1 + x n x s x 1 + x n x x 1 + x n x s+1 x 1, which gives us 3.2 x 1 + x n x = x 1 + x n w i x i [x 1, x n ]. Page 21 of 27

We use 1.10 to pove the theoem. Fo this, we define the n + 2-tuples ξ, η and p as follows: η 1 = x n, η 2 = x n, η 3 = x n 1,..., η n = x 2, η n+1 = x 1, η n+2 = x 1, p 1 = 1, p 2 = w n, p 3 = w n 1,..., p n = w 2, p n+1 = w 1, p n+2 = 1, n+2 ξ 1 = ξ 2 = = ξ n+2 = η, η = p i η i = x 1 + x n w j x j. j=1 It is easily veified that ξ and η ae deceasing and that n+2 p i = 1. It emains to see that ξ, η and p satisfy conditions 1.8 and 1.9. Condition 1.9 is tivially fulfilled since n+2 n+2 n+2 p i ξ i = η p i = η = p i η i. Futhe, we have ξ i = η, i = 1, 2,..., n + 2. To pove 1.8, we need to demonstate that 3.3 η p i p i η i, k = 1, 2,..., n + 1. Fo k = 1, 3.3 becomes η x n, and this holds because of 3.2. On the othe hand, fo k = n + 1, 3.3 becomes η 1 w i x n w i x i, Page 22 of 27

that is, 0 x n x, and this holds because of 3.2. If k {2,..., n}, 3.3 can be ewitten and in its stead we have to pove that 3.4 η 1 w i x n w i x i. i=n+2 k Let us conside the deceasing n-tuple y, whee We have y = = i=n+2 k y i = x 1 + x n x i, i = 1, 2,..., n. w i y i w i x 1 + x n x i = x 1 + x n w i x i = x 1 + x n x = η. If we apply Lemma 3.1 to the n-tuple y and to the weights w, then m = n and fo all l {1, 2,..., n 1} the inequality y l w i l w i x 1 + x n x i Page 23 of 27

holds. Taking into consideation that y = η, l w i = 1 n i=l+1 w i and changing indices as l = n + 1 k, we deduce that 3.5 η 1 i=n+2 k w i n+1 k w i x 1 + x n x i, fo all k {2,..., n}. The diffeence between the ight side of 3.4 and the ight side of 3.5 is n+1 k x n w i x i w i x 1 + x n x i i=n+2 k = x n = x n 1 = x n = i=n+2 k n+1 k i=n+2 k i=n+2 k w i n+1 k w i x i x n w i i=n+2 k i=n+2 k n+1 k w i x n x i + w i x i w i w i x i n+1 k n+1 k n+1 k w i x 1 x i w i x 1 x i w i x i x 1 0, since w is nonnegative and x is inceasing. Theefoe, the inequality n+1 k 3.6 w i x 1 + x n x i x n w i x i i=n+2 k w i x 1 x i Page 24 of 27

holds fo all k {2,..., n}. Fom 3.5 and 3.6 we obtain 3.4. This completes the poof that the m-tuples ξ, η and p satisfy conditions 1.8 and 1.9 and we can apply Theoem E to obtain n+2 p i f η f x n w i f x i + f x 1. Taking into consideation that n+2 p i = 1 and η = x 1 + x n n j=1 w jx j we finally get f x 1 + x n w i x i f x 1 + f x n w i f x i. Page 25 of 27

Refeences [1] S. ABRAMOVICH, G. JAMESON AND G. SINNAMON, Refining Jensen s inequality, Bull. Math. Soc. Math. Roum., 47 2004, 3 14. [2] S. ABRAMOVICH, G. JAMESON AND G. SINNAMON, Inequalities fo aveages of convex and supequadatic functions, J. Inequal. in Pue and Appl. Math., 54 2004, At. 91. [ONLINE: http://jipam.vu.edu. au/aticle.php?sid=444]. [3] S. ABRAMOVICH, S. BANIĆ, M. MATIĆ AND J. PEČARIĆ, Jensen- Steffensen s and elated inequalities fo supequadatic functions, submitted fo publication. [4] S. ABRAMOVICH, M. KLARIČIĆ BAKULA, M. MATIĆ AND J. PEČARIĆ, A vaiant of Jensen-Steffensen s inequality and quasi-aithmetic means, J. Math. Anal. Applics., 307 2005, 370 385. [5] A. McD. MERCER, A monotonicity popety of powe means, J. Inequal. in Pue and Appl. Math., 33 2002, At. 40. [ONLINE: http: //jipam.vu.edu.au/aticle.php?sid=192]. [6] A. McD. MERCER, A vaiant of Jensen s inequality, J. Inequal. in Pue and Appl. Math., 44 2003, At. 73. [ONLINE: http://jipam.vu. edu.au/aticle.php?sid=314]. [7] J.E. PEČARIĆ, F. PROSCHAN AND Y.L. TONG, Convex, Patial Odeings, and Statistical Applications, Academic Pess, Inc. 1992. Page 26 of 27

[8] A. WITKOWSKI, A new poof of the monotonicity popety of powe means, J. Inequal. in Pue and Appl. Math., 53 2004, At. 73. [ONLINE: http://jipam.vu.edu.au/aticle.php?sid=425]. Page 27 of 27