Let \(X\) represent some quantity which is estimated from
data. Let \(\Sigma\) be the (known or estimated)
variance-covariance matrix of \(X\). If \(Y\)
is some computed function of \(X\), then, by the
Delta method (which is a first order Taylor approximation),
the variance-covariance matrix of \(Y\) is approximately
$$\frac{\mathrm{d}Y}{\mathrm{d}{X}} \Sigma \left(\frac{\mathrm{d}Y}{\mathrm{d}{X}}\right)^{\top},$$
where the derivatives are defined over the 'unrolled' (or vectorized)
\(Y\) and \(X\).
Note that \(Y\) can represent a multidimensional quantity. Its
variance covariance matrix, however, is two dimensional, as it too
is defined over the 'unrolled' \(Y\).