Memory-optimized versions of VAR, STD, etc...

From various tests, it appears that

 var(X,0,dim);
 std(X,0,dim);

and maybe other stats functions create a temporary duplicate of X during its calculations. This should be unnecessary (for example, mean and sum do not) and consumes a lot of extra memory when working with large X.

Does anyone know of memory-optimized implementations of these, and maybe other stats functions? I've had no luck with my FEX searches.