How to create a signal matrix

Hello, I am trying to replicate a matrix (signals2) that gives me signals of 1,0 or 1 depending on the value of each single observation in another matrix (rankfull2)-the two have the same size. I would like matrix "signals2" to show 1 when the value is above 319, -1 when is below 60 and 0 otherwise.

 for iii=1:size(rankfull2,1)
 for jjj= 1:size(rankfull2(1,:),2)
    if rankfull2(iii,jjj) < 60
      signals2(iii,jjj)= -1;
    elseif rankfull2(1,jjj) > 319
      signals2(iii,jjj)= 1;
    else signals2(iii,jjj)= 0;
    end
end
end

The output that I get seems to understand the rule of lower than 60 but not the one above 319, returning a 0 instead of a 1 (each row contains at leat 50 values above 319)

Could you someone advise if there is something wrong with my loop or if I am unaware of certain matlab rules?

Thank you in advance.

Answer 1

[David Barry]

It looks like the issue might be because you are only looking in the first row of rankfull2 with your >319 line of code. Anyway, there is a much cleaner way of doing this without the need for loops. You should take a look at logical indexing.

signals2 = zeros(size(rankfull2));
signals2(rankfull2 < 60) = -1;
signals2(rankfull2 > 319) = 1;

[Tommaso Fornaciari]

Great, I am only now getting to know matlab. In my rankfull2 matrix I have some NaNs that change from row to row as well. So I would like the decision criteria not to be fixed but dependent on the number of non NaNs that i have in each row. In other words, immagine the first rankfull2 row to have 600 observations of which 200 NaNs, I would like signals2 to show -1 when rankfull2 is below 40 and 1 when is above 560. The second row will have other criteria depending on the number of non NaNs in rankfull2 row 2 and so on. The criteria is -1 for values below 10% of non NaNs and 1 for values above 90% of non NaNs, allowing me to have the same number of 1s and -1s. Hope I was clear enought.Can I easily adjust the code you provided me with my situation?

[David Barry]

I don't completely follow what you need to do but it sounds like you might need to make use of the insan function along with the logical indexing examples I provided. You could also use the sum function to count the number of NaN in each column if that is what you need.

Take a look at what this example produces:

a = [1, 2, NaN, 4; NaN, 6, 7, 8; NaN, 10, 11, 12];
idx = isnan(a)
sum(idx)

[Tommaso Fornaciari]

Thank you, I have tried to figure it out but I still have problems. I have a matrix C that counts the non NaNs of A per row.

A
  %  1     7    -2     9     4    18    11   NaN    -5    10
 NaN     8    -3     4    22   -58   NaN   NaN     0     3
   6     8    -1     3   NaN    99     2   -50   NaN    43

C
   9
   7
   8

The desired output matrix B would be:

Row1: -1 if lower than 0.1*9 and 1 if bigger than(1-0.1)*9 Row2: -1 if lower than 0.1*7 and 1 if bigger than (1-0.1)*7 Row 3: -1 if lower than 0.1* 8 and 1 if bigger than (1-0.1)*8

Hope you understood. Would greatly appreciate your help again. Thank you.