Create an array with only the increasing values of a pressure time series.

Hi, I'm trying to find a more "elegant" way to create an array using only the increasing values of a time series, so the generated time series will have a positive and increasing slope.

I will appreciate any help, best regards.

clear all,close all;

load DATA;

for i = 1:20; % just a big random number for the if cycle.

   if find(diff(a) < 0);
       id1 = diff(a);
       id1 = find(id1>0);
       a = a(id1);,clear id1;
       
   else
       disp('No more decreasing values')
   end

end

[Dyuman Joshi]

Why are you clearing i? I don't understand that.

And why are you using a loop in the first place? No variable is varying w.r.t the loop.

[Cl4udio]

Hi Dyuman, for is just pushing the if again and again, u can do the same with elseif (running diff only 1 time, will not remove all decreasing values of the time serie, so is running n times)...but is faster like that. Try to run the code with the data. Best regards!

[Dyuman Joshi]

Why not just sort?

Do you want to find the longest non-continuous sub-array that is strictly increasing?

[Cl4udio]

Hi, so imagine that you are on a bout/vessel, and you measure the presurre with a sensor that you attach to a line and , you throw that sensor into the ocean, the bout is moving up and down (ocan waves). So you dont want those re-sampled measures of the ocean during time when the pressure is not increasing, when the instrument is going up and not down. So the sensor is also measuring other variables, so im using the increasing pressure data to uderstand the vertical high frecuency of the ocean....sorry if is confussing.

YES, THOSE ARE THE PERFECT WORDS Longest non-continuous sub-array that is strictly increasing, Thanks

Answer 1

[Torsten]

I prefer a clean for-loop:

a_monotone(1) = a(1);
for i = 2:numel(a)
  if a(i) > a_monotone(end)
    a_monotone(end+1) = a(i);
  else
    a_monotone(end+1) = a_monotone(end);
  end
end
hold on
plot(a_monotone)
plot(a)
hold off

Answer 2

[Voss]

This?

load DATA
plot(a)

idx = diff(a) <= 0;
while any(idx)
    a([false; idx]) = [];
    idx = diff(a) <= 0;
end
plot(a)

Answer 3

[Mathieu NOE]

maybe this ?

I don't see the benefit of removing the negative slope segments , this creates gaps in the signal

was just smoothing or a first order polynomial fit not the best options ?

load DATA;
x = 1:numel(a);

ind_all = [];
val_all = [];
        
for k = 2:numel(a)
    
    d = a(k) - a(k-1);
    
    if d>0 % keep both y(k) - y(k-1)
        ind = [k-1; k];
        val = [a(k-1);  a(k)];
        % store all occurences
        ind_all = [ind_all; ind];
        val_all = [val_all; val];        
    end
    
end
    
% remove duplicates (use ind_all to do this task)
[C,IA,IC] = unique(ind_all);
val_unique_pos = val_all(IA);

plot(x,a,'-*b',C,val_unique_pos,'*d')

[Mathieu NOE]

if you are not too strict about some points at the edges , you can use a one line code with gradient

differences are minor with previous code

% comparison with gradient 
id = find(gradient(a)>0);

plot(x,a,'-*b',C,val_unique_pos,'*r',id,a(id),'*g')

[Cl4udio]

gradient! Look like is the way, dont understand why those green points in the red circlure, those are lower values that the indicaated with the arrows.. Thanks Mathieu! *Gaps are great, all those values that r not in increasing order are removed from the time serie.

notice that

diff(a(gradient(a)>0))

found decreasing values.

, right?