Suppose I have a random variable $X$ and $n$ realizations of this variable: $x_1, ..., x_n$. It seems clear to me in that case that if I am interested in knowing the variability I have in my data (realizations) then I should calculate the variance of $X$ i.e. $var(X)=E[(X-E(X))^2]$ where if I'm interested in its value I can compute its estimate (unbiased sample variance for example): $\frac{1}{n-1}\sum_{i=1}^n(x_i-\hat{\mu})^2$ where $\hat{\mu}=\frac{1}{n}\sum_{i=1}^n x_i$ is the sample mean.

Suppose now that I just take the realizations $x_1, ..., x_n$ and put them inside a vector $\mathbf{x}$. Then if I'm again interested in knowing the variability I have in my data then what should I compute?

  • Is it $var(\mathbf{x})$ and what does it give in that case?
  • Is it a covariance matrix I need to compute i.e. $E[(\mathbf{x}-E(\mathbf{x}))(\mathbf{x}-E(\mathbf{x}))^H]$? If so why?

1 Answer 1


Let $X$ be univariate random variable, say height of an individual. Then $x_{1},x_{2}\cdots,x_{n}$ be a realization on the variable height, and variance is what we compute.

In multivariate analysis, the components of the random vector $X$ are different variables. In this case we will e computing to know how the different components related to each other.

Suppose our random vector is $X=\left(\begin{array}{c} X_{1}\\ X_{2} \end{array}\right)$ where, $X_{1}$ is height, and $X_{2}$ is weight of an individual.

Then, a realization of $n$ values on $X$ will be $\left(\begin{array}{c} x_{11}\\ x_{21} \end{array}\right)$ $\left(\begin{array}{c} x_{12}\\ x_{22} \end{array}\right)$ $\left(\begin{array}{c} x_{13}\\ x_{23} \end{array}\right)$ $\cdots$ $\left(\begin{array}{c} x_{1n}\\ x_{2n} \end{array}\right)$ where the first component of each data value corresponds to height and the second component corresponds to weight. For such data, we compute covariance matrix.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .