This idea is important in the method of moments when one
calculates several moments from the same distribution of events.
In fact, one usually drops (forgets) off diagonal terms in
M and forms from non-independent random variables.
This is done in high energy physics when fitting the moments
, , of a spin particle decay.
Of course, in this case, the off-diagonal elements are never
given by the experimentalist recording his or her data: so one is
forced to set them zero.