The commented-out line deleting the row will, as James noted, break as the array will then have a size(C1,2) one less than the original length (which, btw, will only be correct to begin with if there are more rows than columns in the array) so you will then "run off" the end later on.
As he also said, to do this in a loop requires traversing from last to first as
to not have the problem. But again (as noted also by James) this will also cause a continual reallocation of memory every time it does occur and will be much less efficient than the illustrated use of any supplied function and logical addressing.
The portion
if find (T(i,j)==1)
c1(i,:)=P(i,:);
is also amenable to logical addressing almost identically as to previous
idx=any(T==1),2);
c1(idx,:)=P(idx,:);
NB: the above two cases are two separate logical index vectors using the same variable over; it is also possible to write the expression entirely w/o the temporary variable but that means writing the logical test expression twice; once on each side of the assignment. I'm not sure if the JIT compiler is smart enough to recognize the duplicate operation and do the comparison only once or not; unfortunately TMW doesn't document behavior well enough to know so writing using the temporary ensures the one-time calculation at perhaps expense of some memory. It is also somewhat easier to see what's actually being done so there's some advantage in that, too, particularly for new users.
