number of elements

I have an array A(1483 rows, 417 columns) with integer values of 1 to 5. I want to know whithin each block 92*92(i:i+91,j:j+91), what is the percentage of each value. I looked at size(), hist(), numel() but I don't think they can do what I want. Could you please help me?

Answer 1

[Walter Roberson]

M = A(i:i+91,j:j+91);
h = histc(M(:), 1:5);
p = h ./ (92*92) * 100;

[Hassan]

thanks a lot Walter. your code works fine.

I wrote the following code but yours is more efficient, shorter and faster.

M = A(i:i+91,j:j+91);

n=numel(M)

n1=numel(find(M==1))

n2=numel(find(M==2))

n3=numel(find(M==3))

n4=numel(find(M==4))

n5=numel(find(M==5))

p1=n1*100/n

p2=n2*100/n

p3=n3*100/n

p4=n4*100/n

p5=n5*100/n

Answer 2

[Andrei Bobrov]

variant the full solutions

A = randi([1 5], 1483,417);

m = 92;
mn = fix(size(A)/m);
Awork = A(1:mn(1)*m,1:mn(2)*m);
A1 = reshape(permute(reshape(Awork,m,mn(1),m,[]),[1 3 2 4]),m^2,[]);
A2 = arrayfun(@(x)histc(A1(:,x),1:5),1:prod(mn),'UniformOutput' , false);
Out = cell2mat(A2)/m^2;

more variant

A = randi([1 5], 1483,417);

m = 92;
mn = fix(size(A)/m);
Awork = A(1:mn(1)*m,1:mn(2)*m);
A1 = reshape(permute(reshape(Awork,m,mn(1),m,[]),[1 3 2 4]),m^2,[]);
Out = cell2mat(arrayfun(@(x)histc(A1(:,x),1:5)/nnz(A1(:,x)),1:prod(mn),'UniformOutput' , false));

[Walter Roberson]

Hmmm, yes -- what to do about the boundaries? 417 certainly isn't a multiple of 92. Andrei's code stays sane by using a temporary matrix that is multiples of 92 in each direction.

[Hassan]

thanks Andrei for the comment. I used your code but I didn't get the same results as I got from Walter's.

I have some zero values in my data that want to be excleded while the percentage of values>0 are calculated. could you please help me?

[Hassan]

in my code, I can use n=numel(find(M~=0)).

[Andrei Bobrov]

understood, working

[Walter Roberson]

In my code, if you have some 0 values that you need to be ignored when calculating percentages, then instead of using

p = h ./ (92*92) * 100;

you should use

p = h ./ sum(h) * 100;

Note though that this will give a vector of nan if all the elements in the block are zero.

[Hassan]

thanks a lot Walter.