[R] lower triangle

[Douglas Bates]

"Wilkinson, Mark" <Mark.Wilkinson at stjude.org> writes:

> I want to compute the lower triangle of a square matrix (optionally, sans
> diagonal).  With for() loops I can do something like this:
> 
> ## 5 by 5 matrix rtn
> for (j in 1:5) {
> 	for (k in 1:j) {
> 		if (j != k) {  ## optional
> 		rtn[j, k] <- my.func(j, k)
> 		}
> }
> }
> 
> 
> I'd like to do this with apply().  Is there some way I can do this kind of
> 'short-circuit'?

It depends.  If you can easily calculate a vector of values of
my.func(j,k) in the order in which the elements of the lower triangle
are stored you can do the assignment as

rtn[lower.tri(rtn)] <- vals

or

rtn[lower.tri(rtn, diag = TRUE)] <- vals


If you need to go through something like a double loop to calculate
the new values then this form of the replacement won't provide much of
an advantage.

See ?lower.tri and the lower.tri function itself.