Difference between revisions of "1989 AIME Problems/Problem 11"
m (typo fixes etc) |
m (→Solution: drop minor details?) |
||
Line 15: | Line 15: | ||
At this point, we introduce the crude estimate{{ref|1}} that <math>k=\frac{121-n}{n-1}</math>, so <math>R(n) = 0</math> and | At this point, we introduce the crude estimate{{ref|1}} that <math>k=\frac{121-n}{n-1}</math>, so <math>R(n) = 0</math> and | ||
<center><math>\begin{align*}2S+2n &= (121-n)\left(2001-\frac{121-n}{n-1}\right)+2n \\ | <center><math>\begin{align*}2S+2n &= (121-n)\left(2001-\frac{121-n}{n-1}\right)+2n \\ | ||
− | &= (120-(n-1))\left(2002-\frac{120}{n-1}\right) | + | &= (120-(n-1))\left(2002-\frac{120}{n-1}\right)</math></center> |
− | where <math>C</math> is some constant which does not affect which <math>n</math> yields a maximum. Expanding | + | where <math>C</math> is some constant which does not affect which <math>n</math> yields a maximum. Expanding (ignoring the constants) and scaling, we wish to minimize the expression <math>5(n-1) + \frac{36}{n-1}</math>. By [[AM-GM]], we have <math>5(n-1) + \frac{36}{n-1} \ge 2\sqrt{5(n-1) \cdot \frac{36}{n-1}}</math>, with equality coming when <math>5(n-1) = \frac{36}{n-1}</math>, so <math>n-1 \approx 3</math>. Substituting this result and some arithmetic gives an answer of <math>\boxed{947}</math>. |
− | |||
− | |||
---- | ---- |
Revision as of 16:41, 11 April 2008
Problem
A sample of 121 integers is given, each between 1 and 1000 inclusive, with repetitions allowed. The sample has a unique mode (most frequent value). Let be the difference between the mode and the arithmetic mean of the sample. What is the largest possible value of ? (For real , is the greatest integer less than or equal to .)
Solution
Let the mode be , which we let appear times. We let the arithmetic mean be , and the sum of the numbers be . Then
D &= \left|M-x\right| = \left|\frac{S+xn}{121}-x\right| = \left|\frac{S}{121}-\left(\frac{121-n}{121}\right)x\right|
\end{align*}$ (Error compiling LaTeX. Unknown error_msg)As is essentially independent of , it follows that we wish to minimize or maximize (in other words, ). Indeed, is symmetric about ; consider replacing all of numbers in the sample with , and the value of remains the same. So, without loss of generality, let . Now, we would like to maximize the quantity
contains numbers that may appear at most times. Therefore, to maximize , we would have appear times, appear times, and so forth. We can thereby represent as the sum of arithmetic series of . We let , so
where denotes the sum of the remaining numbers, namely .
At this point, we introduce the crude estimate[1] that , so and
where is some constant which does not affect which yields a maximum. Expanding (ignoring the constants) and scaling, we wish to minimize the expression . By AM-GM, we have , with equality coming when , so . Substituting this result and some arithmetic gives an answer of .
In less formal language, it quickly becomes clear after some trial and error that in our sample, there will be values equal to one and values each of . It is fairly easy to find the maximum. Try , which yields , , which yields , , which yields , and , which yields . The maximum difference occurred at , so the answer is .
Notes
- ^ In fact, when (which some simple testing shows that the maximum will occur around), it turns out that is an integer anyway, so indeed .
See also
1989 AIME (Problems • Answer Key • Resources) | ||
Preceded by Problem 10 |
Followed by Problem 12 | |
1 • 2 • 3 • 4 • 5 • 6 • 7 • 8 • 9 • 10 • 11 • 12 • 13 • 14 • 15 | ||
All AIME Problems and Solutions |