1.7 Variance and Standard Deviation

Another important quantity related to a given random variable is its variance. The variance is a numerical description of the spread, or the dispersion, of the random variable. That is, the variance of a random variable X is a measure of how spread out the values of X are, given how likely each value is to be observed.

Variance and Standard Deviation of a Discrete Random Variable
The variance, Var(X), of a discrete random variable X is ${\text{Var}}(X)=\sum _{k=1}^{N}{\Big (}x_{k}-\mathbb {E} (X){\Big )}^{2}{\textrm {Pr}}(X=x_{k})$ where N is the total number of possible values of X. The standard deviation, σ, is the positive square root of the variance: $\sigma (X)={\sqrt {{\text{Var}}(X)}}$

Observe that the variance of a random variable is always nonnegative (since probabilities are nonnegative, and the square of a number is also nonnegative).

Observe also that much like the expectation of a random variable X, the variance (or standard deviation) is a weighted average of an expression of observable and calculable values. More precisely, notice that

${\text{Var}}(X)=\mathbb {E} \left(\left[X-\mathbb {E} (X)\right]^{2}\right).$

Example: Test Scores

Using the test scores example of the previous sections, calculate the variance and standard deviation of the random variable X associated to randomly selecting a single exam.

Solution

The variance of the random variable X is given by

${\begin{aligned}{\text{Var}}(X)&=\sum _{k=1}^{N}(x_{k}-\mathbb {E} (X))^{2}{\textrm {Pr}}(X=x_{k})\\&=(30-64)^{2}{\frac {3}{10}}+(60-64)^{2}{\frac {2}{10}}+(80-64)^{2}{\frac {3}{10}}+(90-64)^{2}{\frac {1}{10}}+(100-64)^{2}{\frac {1}{10}}\\&=624\end{aligned}}$

The standard deviation of X is then

$\sigma (X)={\sqrt {624}}\approx 24.979992$

Interpretation of the Standard Deviation

For most "nice" random variables, i.e. ones that are not too wildly distributed, the standard deviation has a convenient informal interpretation. Consider the intervals

$S_{m}=\left[\mathbb {E} (X)-m\sigma (X),\ \mathbb {E} (X)+m\sigma (X)\right],$

for some positive integer m. As we increase the value of m, these intervals will contain more of the possible values of the random variable X.

A good rule of thumb is that for "nicely distributed" random variables, all of the most likely possible values of the random variable will be contained in the interval S₃. Another way to say this is that, for discrete random variables, most of the PMF will live on the interval S₃. We will see in the next chapter that a similar interpretation holds for continuous random variables.