Can zero rows in matrices be ignored in calculations of matrix products? I understand that when calculating the product of 2 matrices you need to account for the dimensions. But when there is an empty row in one of the matrices, why does it need to be accounted for? What are the implications if you ignore that row before performing a calculation? 
Say, if the matrix is 
$\begin{bmatrix}0&-2\\
-1&4\\
0&0\\ \end{bmatrix},$
in which ways is it different if the last row was removed?
 A: Zero rows or columns can be ignored
if they are in the proper orientation
(i.e., rows if the matrix
is on the left
and columne if the matrix is on the right
of the product).
If you know in advance that
a matrix has (or is likely to have)
at least one zero row
(or column),
it is probably useful
to find all such items
and modify your loop
to take them into account.
Of course,
if your hardware makes it much more efficient
to do every term in every product,
don't bother.
A: To put it shortly: It cannot be ignored.
A matrix $A$ is not just a bag full of numbers, but is organized information of some sort. An essential part of this organization consists in the number $m$ of rows and the number $n$ of columns of $A$. The rows are in a prearranged $1$-$1$ correspondence with the elements of some set of cardinality $m$ (say, a set of $m$ cities), and the columns are in a $1$-$1$ correspondence with the elements of some other set of cardinality $n$ (say $n$ kinds of crimes). The entry $a_{ik}$ of $A$ might be the number of crimes of kind$_k$  that have occurred in city$_i$ in 2013. If row$_i$ of $A$ is a row of zeros then this is a substantial fact about city$_i\,$, which definitively should not be ignored but carried through the whole statistical, resp. computational,  analysis of the data.
A: When multiplying, rows with 0's have to be considered because if not, the order of the matrices will not be defined.
