Find Matrix Elements That Meet a Condition

I have a 1025x36 single matrix A and I like to find specific matrix elements by having the following condition fulfilled:

B = A( A>4 & A<8);

After finding all the values that meet this condition, I would like these values to be averaged columnwise for each of the 36 columns.

Can somebody help how to do this. When I find the values that meet the condition, they all get concatenated vertically and I am unable to calculate the mean for these values per column.

Answer 1

[KSSV]

You can pick a particular column and run a loop.

    for i = 1:36
        C = A(:,i) ;
        B = C( C>4 & C<8);
        %%Do what you want 
    end

[Stephan Raczak]

Thank you!

Answer 2

[Stephen23]

Although there are many ways of doing this, here are two robust methods. Note that using a loop might seem attractive to beginners, but is not a very effective use of MATLAB's abilities, and will not scale well to larger array sizes.

Method One: DIY: add zeros in the non-matching locations, and then calculate the mean yourself:

>> A = randi(9,7,10)
  A =
       8     5     3     3     6     7     5     9     7     9
       4     5     7     7     4     3     2     3     5     2
       1     8     3     8     9     8     4     7     4     7
       3     5     7     4     1     5     6     2     6     3
       2     9     7     8     5     1     2     3     6     2
       3     6     1     7     4     2     7     1     7     6
       4     9     3     1     5     7     3     6     6     5
>> B = A;
>> idx = A>4 & A<8;
>> B(~idx) = 0;
>> sum(B,1) ./ sum(idx,1)
ans =
       NaN    5.2500    7.0000    7.0000    5.3333    6.3333    6.0000    6.5000    6.1667    6.0000

Method Two: use nanmean. Add NaN in the non-matching locations, and then use the nanmean function (which ignores the NaNs):

>> B = A;
>> B(~(A>4 & A<8)) = NaN;
>> nanmean(B,1)
ans =
       NaN    5.2500    7.0000    7.0000    5.3333    6.3333    6.0000    6.5000    6.1667    6.0000

If your MATLAB version does not have nanmean, then you can find some versions on File Exchange, e.g.:

https://www.mathworks.com/matlabcentral/fileexchange/6837-nan-suite