Author 
Message 
Intern
Joined: 30 Aug 2010
Posts: 10

This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 12:18
Question Stats:
0% (00:00) correct
100% (02:14) wrong based on 1 sessions
HideShow timer Statistics
Given a list of known arithmetic mean and standard deviation \(x_1,x_2,...,x_n\) find a number \(x_n_+_1\) such that it will not change the standard deviation of the list. I have the solution, but because I am new to the forum I have to wait five days before I can give the website.



Manager
Joined: 25 Jun 2010
Posts: 91

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 12:51
X(n+1) = (X1+X2+...+Xn)/n (means the mean)?
Last edited by anshumishra on 30 Aug 2010, 13:08, edited 1 time in total.



CEO
Status: Nothing comes easy: neither do I want.
Joined: 12 Oct 2009
Posts: 2784
Location: Malaysia
Concentration: Technology, Entrepreneurship
GMAT 1: 670 Q49 V31 GMAT 2: 710 Q50 V35

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 12:57
trx123 wrote: Given a list of known arithmetic mean and standard deviation \(x_1,x_2,...,x_n\) find a number \(x_n_+_1\) such that it will not change the standard deviation of the list. I have the solution, but because I am new to the forum I have to wait five days before I can give the website. If the difference between the new element and the arithmetic mean is equal to the standard deviation then the deviation wont be changed by the addition of new entry.
_________________
Fight for your dreams :For all those who fear from Verbal lets give it a fight
Money Saved is the Money Earned
Jo Bole So Nihaal , Sat Shri Akaal
Support GMAT Club by putting a GMAT Club badge on your blog/Facebook
GMAT Club Premium Membership  big benefits and savings
Gmat test review : http://gmatclub.com/forum/670to710alongjourneywithoutdestinationstillhappy141642.html



Director
Status: Apply  Last Chance
Affiliations: IIT, Purdue, PhD, TauBetaPi
Joined: 18 Jul 2010
Posts: 685
Schools: Wharton, Sloan, Chicago, Haas
WE 1: 8 years in Oil&Gas

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 12:59
The current mean should do it.. Posted from my mobile device
_________________
Consider kudos, they are good for health



Intern
Joined: 30 Aug 2010
Posts: 10

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 15:04
gurpreetsingh, your answer is approximately correct only for large n. The exact answer is: \(x_n_+_1=\mu_n\pm\sigma_n\sqrt{\frac{n+1}{n}}\) The derivation of this formula is not straightforward and took me quite some time. I have the derivation, but unfortunately cannot post my website address yet because I just joined the forum today.



Intern
Joined: 30 Aug 2010
Posts: 10

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 15:11
mainhoon, Sorry, no cigar. the mean will not change the mean, but it will change the standard deviation.



CEO
Status: Nothing comes easy: neither do I want.
Joined: 12 Oct 2009
Posts: 2784
Location: Malaysia
Concentration: Technology, Entrepreneurship
GMAT 1: 670 Q49 V31 GMAT 2: 710 Q50 V35

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 15:21
Let d= standard deviation and a = Mean \(d = \sqrt{\frac{{(ax_1)^2 + (ax_2)^2 + (ax_2)^2 .....}}{n}}\) New \(d_n\)= \(\sqrt{\frac{{(ax_1)^2 + (ax_2)^2 + (ax_2)^2 .....(ax_{n+1})^2}}{{n+1}}}\) For \(d = d_n\) \({d_n}^2 = \frac{{(ax_1)^2 + (ax_2)^2 + (ax_2)^2 .....(ax_{n+1})^2}}{{n+1}}\) \({d}^2 = \frac{{ d^2 *n+ (ax_{n+1})^2}}{{n+1}}\) \({d}^2 * (n+1) = { d^2 *n+ (ax_{n+1})^2}\) => \({d}^2 = {(ax_{n+1})^2}\) => \(d = ax_{n+1}\)
_________________
Fight for your dreams :For all those who fear from Verbal lets give it a fight
Money Saved is the Money Earned
Jo Bole So Nihaal , Sat Shri Akaal
Support GMAT Club by putting a GMAT Club badge on your blog/Facebook
GMAT Club Premium Membership  big benefits and savings
Gmat test review : http://gmatclub.com/forum/670to710alongjourneywithoutdestinationstillhappy141642.html
Last edited by gurpreetsingh on 30 Aug 2010, 15:24, edited 1 time in total.



Director
Status: Apply  Last Chance
Affiliations: IIT, Purdue, PhD, TauBetaPi
Joined: 18 Jul 2010
Posts: 685
Schools: Wharton, Sloan, Chicago, Haas
WE 1: 8 years in Oil&Gas

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 15:23
But why is the mean in the n and n+1 case the same? a? that should change...I think the idea is to equate both the standard deviation formula and calculate what the new value should look like.. But I doubt if this is GMAT type...
_________________
Consider kudos, they are good for health



Intern
Joined: 30 Aug 2010
Posts: 10

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 15:59
gurpreetsingh, Your derivation is surprisingly similar to my first attempt, and it happens to be correct as an approximation only for large n. But there is a flaw in this approach, because you are assuming that the mean is the same in both lists, which is not true.
mainhoon, You are right, the mean will change, and that is the flaw in gurpreetsingh derivation. The derivation of the exact formula is quite involved. Also, I am sure this is not a GMAT question.



Intern
Joined: 30 Aug 2010
Posts: 10

Re: This problem is not easy [#permalink]
Show Tags
30 Aug 2010, 17:32
The Problem: Find a number, such that when added to a list of numbers will not change the standard deviation of the original list.
The solution: Define the list of numbers as: \($x_1 ,x_2 , \cdots x_n ,x_{n + 1}$\) where \($x_{n + 1} $\) is the number added to the list. The mean and the standard deviation of the original list are: \(\bar x_n = \frac{1} {n}\sum\limits_{i = 1}^n {x_i } \\) \(\sigma _n ^2 = \frac{1} {{n  1}}\sum\limits_{i = 1}^n {\left( {x_i  \bar x_n } \right)^2 } \\) After \($x_{n + 1} $\) is added to the original list, the mean and the standard deviation of the new list are: \(\bar x_{n + 1} = \frac{1} {{n + 1}}\sum\limits_{i = 1}^{n + 1} {x_i } \\) \(\sigma _{n + 1} ^2 = \frac{1} {n}\sum\limits_{i = 1}^{n + 1} {\left( {x_i  \bar x_{n + 1} } \right)^2 } \\)
We want to find \($x_{n + 1} $\)
such that \($\sigma _n ^2 = \sigma _{n + 1} ^2 $\)
therefore, \(\frac{1} {{n  1}}\sum\limits_{i = 1}^n {\left( {x_i  \bar x_n } \right)^2 } = \frac{1} {n}\sum\limits_{i = 1}^{n + 1} {\left( {x_i  \bar x_{n + 1} } \right)^2 } \\) The sum of terms up to n is \(\sum\limits_{i = 1}^n {x_i = n\bar x_n } \\) and the sum of terms up to n+1 is \(\sum\limits_{i = 1}^{n + 1} {x_i = (n + 1)\bar x_{n + 1} } \\) Subtracting \($x_{n + 1}$\)
\(\sum\limits_{i = 1}^{n + 1} {x_i  x_{n + 1} = \sum\limits_{i = 1}^n {x_i } } = n\bar x_n \\) therefore,
\(\(n + 1)\bar x_{n + 1}  x_{n + 1} = n\bar x_n \\)
\(\bar x_{n + 1} = \frac{n} {{n + 1}}\bar x_n + \frac{{x_{n + 1} }} {{n + 1}} \\) (1)
Some algebra:
\(\sum\limits_{i = 1}^{n + 1} {\left( {x_i  \bar x_{n + 1} } \right)^2 } = \left( {x_{n + 1}  \bar x_{n + 1} } \right)^2 + \sum\limits_{i = 1}^n {\left( {x_i  \bar x_{n + 1} } \right)^2 } \\) Substituting \(\bar x_{n + 1} \\) with (1) \(\sum\limits_{i = 1}^{n + 1} {\left( {x_i  \bar x_{n + 1} } \right)^2 } = \left( {x_{n + 1}  \frac{n} {{n + 1}}\bar x_n  \frac{{x_{n + 1} }} {{n + 1}}} \right)^2 + \sum\limits_{i = 1}^n {\left( {x_i  \frac{n} {{n + 1}}\bar x_n  \frac{{x_{n + 1} }} {{n + 1}}} \right)^2 } \\) Rearranging, \(\ = \left( {\frac{n} {{n + 1}}x_{n + 1}  \frac{n} {{n + 1}}\bar x_n } \right)^2 + \sum\limits_{i = 1}^n {\left( {x_i  \frac{n} {{n + 1}}\bar x_n  \frac{{x_{n + 1} }} {{n + 1}}} \right)^2 } \\) \(\ = \left( {\frac{n} {{n + 1}}(x_{n + 1}  \bar x_n )} \right)^2 + \sum\limits_{i = 1}^n {\left( {x_i  \bar x_n + \frac{{\bar x_n }} {{n + 1}}  \frac{{x_{n + 1} }} {{n + 1}}} \right)^2 } \\) \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {\left( {x_i  \bar x_n + \frac{1} {{n + 1}}(\bar x_n  x_{n + 1} )} \right)^2 } \\) Expanding the square in the summation, \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {\left( {(x_i  \bar x_n )^2 + \frac{{2(x_i  \bar x_n )}} {{n + 1}}(\bar x_n  x_{n + 1} ) + \frac{1} {{(n + 1)^2 }}(\bar x_n  x_{n + 1} )^2 } \right)} \\) Note that \(\ (x_{n + 1}  \bar x_n )^2 = (\bar x_n  x_{n + 1} )^2 \\) Therefore \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {\left( {(x_i  \bar x_n )^2 + \frac{{2(x_i  \bar x_n )}} {{n + 1}}(\bar x_n  x_{n + 1} ) + \frac{1} {{(n + 1)^2 }}(x_{n + 1}  \bar x_n )^2 } \right)} \\) If C is a constant \(\sum\limits_{i = 1}^n {(f(i) + C) = } \sum\limits_{i = 1}^n {f(i) + nC} \\) \(\sum\limits_{i = 1}^n {f(i) + Cg(i) = } \sum\limits_{i = 1}^n {f(i) + C\sum\limits_{i = 1}^n {g(i)} } \\) Therefore \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 + \sum\limits_{i + 1}^n {\left( {\frac{{2(x_i  \bar x_n )}} {{n + 1}}(\bar x_n  x_{n + 1} ) + \frac{1} {{(n + 1)^2 }}(x_{n + 1}  \bar x_n )^2 } \right)} } \\) \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 + \frac{2} {{n + 1}}(\bar x_n  x_{n + 1} )\sum\limits_{i + 1}^n {(x_i  \bar x_n ) + \frac{n} {{(n + 1)^2 }}(x_{n + 1}  \bar x_n )^2 } } \\) Observe that \(\sum\limits_{i = 1}^n {(x_i  \bar x_n )} = \sum\limits_{i = 1}^n {x_i }  n\bar x_n = n\bar x_n  n\bar x_n = 0 \\) Therefore \(\ = \left( {\frac{n} {{n + 1}}} \right)^2 (x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 + } \frac{n} {{(n + 1)^2 }}(x_{n + 1}  \bar x_n )^2 \\) \(\ = \frac{n} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 } \\) From the original problem, \(\frac{1} {{n  1}}\sum\limits_{i = 1}^n {\left( {x_i  \bar x_n } \right)^2 } = \frac{1} {n}\sum\limits_{i = 1}^{n + 1} {\left( {x_i  \bar x_{n + 1} } \right)^2 } \\) Therefore, we need \(x_{n + 1} \\) to satisfy \(\frac{1} {{n  1}}\sum\limits_{i = 1}^n {\left( {x_i  \bar x_n } \right)^2 } = \frac{1} {n}\left( {\frac{n} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 + \sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 } } \right) = \frac{1} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 + \frac{1} {n}\sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 } \\) \(\sigma _n ^2 = \frac{1} {{n  1}}\sum\limits_{i = 1}^n {\left( {x_i  \bar x_n } \right)^2 } = \frac{1} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 + \frac{1} {n}\sum\limits_{i = 1}^n {(x_i  \bar x_n )^2 } \\) \(\sigma _n ^2 = \frac{1} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 + \frac{{n  1}} {n}\sigma _n ^2 \\) \(\sigma _n ^2  \frac{{n  1}} {n}\sigma _n ^2 = \frac{1} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 \\) \(\frac{1} {n}\sigma _n ^2 = \frac{1} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 \\) \(\sigma _n ^2 = \frac{n} {{(n + 1)}}(x_{n + 1}  \bar x_n )^2 \\) \(x_{n + 1} = \bar x_n \pm \sigma _n \sqrt {\frac{{n + 1}} {n}}\\) For large n \(x_{n + 1} \approx \bar x_n \pm \sigma _n \\) For large n, adding an observation one standard deviation on either side of the mean will not change the standard deviation significantly.




Re: This problem is not easy
[#permalink]
30 Aug 2010, 17:32






