Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $a_1,$ $a_2,$ $\ldots,$ $a_n$ be positive real numbers. Prove that $$\sqrt{\frac{a_1^2+\left( \frac{a_1+a_2}{2}\right)^2+\cdots +\left(\frac{a_1+a_2+\cdots +a_n}{n}\right)^2}{n}} \le \frac{a_1+\sqrt{\frac{a_1^2+a_2^2}{2}}+\cdots+\sqrt{\frac{a_1^2+a_2^2+\cdots +a_n^2}{n}}}{n}.$$

I have proved this inequality for $n=2$ and $n=3.$ But I still cannot prove it for the general case. Can somone help me?

share|improve this question
    
I came across a monthly article by a graduate student, which has a strong resemblance to the one you ask. but I cannot remember which one. –  Sunni May 19 '10 at 16:58
2  
You'll probably have more luck at the art of problem solving forums artofproblemsolving.com/Forum/index.php –  j.c. May 19 '10 at 18:42
2  
I like the question and am tempted to have a go -- but no time right now. –  gowers May 19 '10 at 22:54

1 Answer 1

up vote 4 down vote accepted

Mixed mean inequalities have been studied quite a bit, inspired mostly by the inequalities of Carleman and Hardy, starting probably from this article of K. Kedlaya. Note that the following holds if $r < s$ (in your case $r=1, s=2$): $$\left(\frac{1}{n}\sum_{k=1}^n \left(M_k^{[r]}(\mathbf a)\right) ^s\right)^{1/s}\leq \left(\frac{1}{n}\sum_{k=1}^n\left(M_k^{[s]}(\mathbf a)\right) ^r \right)^{1/r}$$ Where $\mathbf a$ denotes a sequence of real numbers, and $M_k^{[r]}$ denotes the $r$-th power mean of the first $k$ variables. You can find a proof of the general form in "Survey on classical inequalities" by T.M. Rassias (page 32, it is the original proof of B. Mond and J. Pecaric which was later extended to matrices and linear operators).

share|improve this answer
    
Thanks Gjergji Zaimi for your reference. –  can_hang2007 Jun 2 '10 at 1:38

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.