Indexing the 2 D matrices with the help of one more dimension

If there is code for a random matrix A of m*n dimension using forward space forward time finite difference method i.e.

A(i+1,j ) = some function of (A(i,j)+rand(i,j))

Since A is now random matrix, so to each different run we will get definitely different matrix. So to resolve this, I want to find at least 1000 A matrix, and then I will take average of all and so the new matrix can be considered as the expected matrix. But I am unable to create Matlab code for this. Please help