Let's say you want to extract the two elements equal to 1 in this matrix.
A = [1 2 3; 4 5 6; 7 8 1]
You can do this with find and ind2sub.
locations = find(A == 1)
[row, col] = ind2sub(size(A), locations)
Now let's extract the 1's from A.
B = A(row, col)
Why are the 3 and 7 in there? Because B contains all the elements at the intersections of rows in the variable row and columns in the variable col. So even though you may have expected B to have only two elements because row and col each had two elements, it actually has two times two or four elements.
Now let's look at a slightly different matrix:
C = [1 1 3; 4 5 6; 7 8 1];
locationsC = find(C == 1);
[rowC, colC] = ind2sub(size(C), locationsC)
D = C(rowC, colC)
These are the elements at the intersections of rows 1, 1, and 3 and columns 1, 2, and 3. Given that knowledge, even though arr is of size (800,800,3,9,9) the array arr(x, y ,3, z ,w) may be much larger.
If you just want the mean of elements in an array corresponding to locations of a specific value in another array, don't use subscripted indexing. Use logical indexing.
justThreeElements = A(C == 1)
mean(justThreeElements)
The fact that you have that other dimension in arr is a little tricky, but you could extract just the third page to a separate variable.
