PlanetMath (more info)
 Math for the people, by the people. Sponsor PlanetMath
Encyclopedia | Requests | Forums | Docs | Wiki | Random | RSS  
Login
create new user
name:
pass:
forget your password?
Main Menu
Revision difference : derivation of geometric mean as the limit of the power mean
Version current Version 4
\PMlinkescapeword{fix} \PMlinkescapeword{fix}
\PMlinkescapeword{calculate} \PMlinkescapeword{calculate}
Fix $x_1, x_2, \ldots, x_n \in \mathbb{R}^+$. Then let Fix $x_1, x_2, \ldots, x_n \in \mathbb{R}^+$. Then let
\[ \[
\mu(r) := \left(\frac{x_1^r+\cdots+x_n^r}{n}\right)^{1/r}. \mu(r) := \left(\frac{x_1^r+\cdots+x_n^r}{n}\right)^{1/r}.
\] \]
For $r\neq 0$, by definition $\mu(r)$ is the $r$th power mean of the $x_i$. It is also clear that $\mu(r)$ is a differentiable function for $r\neq 0$. What is $\lim_{r\to 0} \mu(r)$? For $r\neq 0$, by definition $\mu(r)$ is the $r$th power mean of the $x_i$. It is also clear that $\mu(r)$ is a differentiable function for $r\neq 0$. What is $\lim_{r\to 0} \mu(r)$?
We will first calculate $\lim_{r\to 0} \log\mu(r)$ using \PMlinkname{l'H\^opital's rule}{LHpitalsRule}. We will first calculate $\lim_{r\to 0} \log\mu(r)$.
\begin{align*} \begin{align*}
\lim_{r\to 0} \log\mu(r) & = \lim_{r\to 0} \frac{\log\left(\frac{x_1^r+\cdots +x_n^r}{n}\right)}{r}\\ \lim_{r\to 0} \log\mu(r) & = \lim_{r\to 0} \frac{\log\left(\frac{x_1^r+\cdots +x_n^r}{n}\right)}{r}\\
\intertext{Note that this is a limit of the form $0/0$ so we can apply \PMlinkname{l'H\^opital's rule.}{LHpitalsRule}}
& = \lim_{r\to 0} \frac{\left(\frac{x_1^r\log x_1+\cdots+x_n^r\log x_n}{n}\right)}{\left(\frac{x_1^r+\cdots+x_n^r}{n}\right)}\\ & = \lim_{r\to 0} \frac{\left(\frac{x_1^r\log x_1+\cdots+x_n^r\log x_n}{n}\right)}{\left(\frac{x_1^r+\cdots+x_n^r}{n}\right)}\\
& = \lim_{r\to 0} \frac{x_1^r\log x_1+\cdots+x_n^r\log x_n}{x_1^r+\cdots+x_n^r}\\ & = \lim_{r\to 0} \frac{x_1^r\log x_1+\cdots+x_n^r\log x_n}{x_1^r+\cdots+x_n^r}\\
& = \frac{\log x_1+\cdots+\log x_n}{n}\\ & = \frac{\log x_1+\cdots+\log x_n}{n}\\
& = \log \sqrt[n]{x_1\cdots x_n}. & = \log \sqrt[n]{x_1\cdots x_n}.
\end{align*} \end{align*}
It follows immediately that It follows immediately that
\[ \[
\lim_{r\to 0} \left(\frac{x_1^r+\cdots+x_n^r}{n}\right)^{1/r} = \sqrt[n]{x_1\cdots x_n}. \lim_{r\to 0} \left(\frac{x_1^r+\cdots+x_n^r}{n}\right)^{1/r} = \sqrt[n]{x_1\cdots x_n}.
\] \]