next up previous contents
Next: Illustrations and Examples Up: Characteristic Parameters of a Previous: Correlations   Contents


Arithmetic Mean of Random Variables

$Z$ denotes a random variable with $m=EZ$ and $/sigma^{2}=Var(Z)=E(Z-m)^{2}.$ Sample values $z_{1},/ldots,z_{n}$ are interpreted as $n$ independent realizations of $Z$. They also can be interpreted as realizations of $n$ independent random variables $Z_{1},/ldots,Z_{n},$ which all have the same distribution as $Z$. This $Z_{i}, /; i=1,/ldots,n,$ are called sample variables. The arithmetic mean

/begin{displaymath}/bar{Z}=/frac{1}{n} /sum_{i=1}^{n} Z_i/end{displaymath}

then is a random variable, with a given distribution, of which we would like to compute the expectation and variance.

From this we need two properties of linear combinations of independent random variables:

/begin{displaymath}E(/sum a_{i}Z_{i})=/sum a_{i}EZ_{i}/ / / ,/end{displaymath}


/begin{displaymath}Var(/sum a_{i}Z_{i})=/sum a_i^2 Var(Z_{i})/ / / ./end{displaymath}

It follows immediately for the arithmetic mean

/begin{displaymath}E /bar{Z}=E[/sum_{i=1}^{n}(/frac{1}{n} Z_{i})]=/sum /frac{1}{n} EZ_{i}=
/frac{1}{n}/sum m = m/end{displaymath}

and

/begin{displaymath}/sigma^2_{/bar{Z}}=Var(/bar{Z})=Var[/sum(/frac{1}{n}Z_{i})]=
...
...2} Var(Z_{i})=/frac{1}{n^2} /sum /sigma^2 = /sigma^{2}/n/
/ / ./end{displaymath}

The corresponding deviation or standard error is

/begin{displaymath}/sigma_{/bar{Z}}=/sigma//sqrt{n}/ / / ,/end{displaymath}

and we summarize:

The arithmetic mean shows the same expectation as the random variable $Z_{i}$ and a standard deviation (or error), that reduces that of $Z_{i}$ by a factor $/sqrt{n}$.

From this, for example, if we assume an approximate normal distribution of the data, we can give immediately an important confidence interval for the theoretical mean value $m$ of the distribution (Hartung et al., 1984[9]): with about 95% probability the interval

/begin{displaymath}/bar{z} /pm 2 s/ /sqrt{n}/end{displaymath}

covers the value $m$, where $/bar{z}$ respectively $s$ denote again the estimated (sample) values of $m$ and $/sigma $.


next up previous contents
Next: Illustrations and Examples Up: Characteristic Parameters of a Previous: Correlations   Contents
Rudolf Dutter 2003-03-13