I would like to filter out some vectors based on a percent of aggregate

I would like to filter some time series based on their weight of the overall aggregate. The weights are to be calculated based on the derived aggregate/components of the second last element in each vector (x1=20, x2=2, x3=62) The end result would be to extract the complete time series vectors satisfying the threshold. I would also appreciate suggested improvements to the code:

Threshold = *.2*

x1=[5 10 20 15 40 20 25]';
x2=[8 25 15 50 41 2 15]';
x3=[32 42 30 52 33 62 77]';
x=[x1 x2 x3];
sumx=sum(x')';
sumx=[sumx sumx sumx];
perx=x./sumx
perx=perx(4,:);

perx =

    0.2381    0.0238    0.7381

Here x2 should be deleted since it falls below the threshold and a new object containing x1 and x3 vectors should get created.

x=[x1 x3];

[Image Analyst]

How does x2(end-1) being 2 fall below a threshold of 0.2? I don't follow the logic of the sumx and perx part. What's going on there?

[Jeff]

Hi Image, you are on the right track thanks for your help. The one thing that was missing is that the threshold is based on the percentage that the second last element makes of the total of the three vectors along that element;

sumx=20+2+62 = 84 x1%=20/84=.2381 x2%=2/84=.0238 x3%=62/84=.7381

since threshold=.2 then x2%<threshold and x2 vector should be dropped;

Is that clearer?

[Image Analyst]

Yes, see my new edited code below.

Answer 1

[Image Analyst]

See if this is what you want:

thresholdPercent = 0.2
x1=[5 10 20 15 40 20 25]';
x2=[8 25 15 50 41 2 15]';
x3=[32 42 30 52 33 62 77]';
% Sum them up
sumx = sum([x1,x2,x3], 2)
x=[];
% Get the value that each column will be compared to.
Threshold = thresholdPercent * sumx(end-1)
% Tack on x1 if it's above the threshold.
if x1(end-1) > Threshold
  fprintf('Adding on x1:\n');
  x = [x, x1]
end
% Tack on x2 if it's above the threshold.
if x2(end-1) > Threshold
  fprintf('Adding on x2:\n');
  x = [x, x2]
end
% Tack on x3 if it's above the threshold.
if x3(end-1) > Threshold
  fprintf('Adding on x3:\n');
  x = [x, x3]
end
% Get the means:
meanAcrossColumns = mean(x, 2)

[Jeff]

Thanks Image that's it!! Very cool; I may need your help again if you don't mind.

Curious could you tell me what this does? fprintf('Adding on x1:\n');

Again much appreciated for that solid code

[Image Analyst]

It just prints out some debugging stuff to the command window - it's optional and you can remove it if you want.

[Jeff]

Hi Imagine, just implemented your code with actual data I have and one question that comes up is how do I know which variable has passed this threshold test? Is there a way to not only store that variable in object x but to name it as well so that I know which variable I have? Much appreciated

Answer 2

[Jeff]

Found this in the matlab documentation just to share

Find the indices of values in a vector that are greater than a specified limit.

A = rand(1,10); limit = .75;

B = (A > limit); % B is a vector of logical values if any(B) fprintf('Indices of values > %4.2f: \n', limit); disp(find(B)) else disp('All values are below the limit.') end