How to calculate the highest consecutive negative results in my code

Hi,

I'm new to Matlab and I have been having help to write this code. I am trying to learn on my own and I am trying now to write something that would calculate the highest consecutive negative results in my code. I would be happy to recieve some help!

clear variables
k = 5000;

risk = 0.007;

riskM = 0.1;

maxR = 0.022;

numberSteps = 105;

markets = zeros(numberSteps,1);
markets(1) = k;
for i = 1:size(markets,1) - 1 
    markets(i+1) = markets(i)*(1+0.1); 
end

% ka = 1100

%%
numberSim = 1000;
distR = randsrc(numberSteps, numberSim,[-1 -.6 -.36 -.07 0 0.17 0.37 0.39 0.5 0.81 1.37 1.39 1.4 2.2 2.57 3.16 3.55 5.49 6.56 7.34 7.82 8.63 8.7 9.15 9.42 11.55 17.94 19.9 21.18 22.28 28.4 28.51;...
                                        .648 .0095 .0095 .0095 0.067 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095]);
cumR = cumsum(distR,1);

MaxD = maxdrawdown(cumR,'arithmetic');
[maxMaxD,ii] = max(MaxD);
meanD = mean(MaxD);
stdD = std(MaxD);

  
figure
subplot(1,2,1)
plot(cumR)
subplot(1,2,2)
plot(cumR(:,ii))
sgtitle(['Mean Drawdown=',num2str(meanD,3),'   std=',num2str(stdD,3), '   Max Possible DD=',num2str(maxMaxD,3)])

[Mohammad Sami]

I am not sure what you mean by highest consecutive negative results.

[Camilo Acevedo]

Hi Mohammad, what I mean is that for instance one of the random distribution is [2 4 3 -1 -2 5 7 8 -1 -0.3 -1 -1 -1 -1 -1 -1 -1 7] an another might be [2 2 2 4 3 -1 -1 -2 4 6 7 8 -3 -2 -1 -0.2 4 6]. So what I need is a code that finds where in all the random distribution is the highest consecutive negative sum (consecutive negative number), which in this example would be -8.3 (-1 -0.3 -1 -1 -1 -1 -1 -1 -1).

Answer 1

[Image Analyst]

This is trivial if you have the Image Processing Toolbox. Simply use regionprops():

v = [2 4 3 -1 -2 5 7 8 -1 -0.3 -1 -1 -1 -1 -1 -1 -1 7] % Sample data.
% Find elements where v is negative.
negativeValues = v < 0
% Measure area and mean of the negative regions.
props = regionprops(negativeValues, v, 'MeanIntensity', 'Area');
% Extract values from structure into arrays, for convenience.
allMeanIntensities = [props.MeanIntensity]
allAreas = [props.Area]
% Get the sum of all values in each region.
integratedValues = allMeanIntensities .* allAreas
% Find the sum that is MOST negative.
mostNegative = min(integratedValues)

[Camilo Acevedo]

Thank you!! It worked great!!

[Camilo Acevedo]

I have a problem with this code. Although it worked fine with the sample of the example you wrote, when I put the random results (which are more than one distribution) gives me values that are incorrect:

Any ideas why?

clear variables
k = 5000;

risk = 0.007;

riskM = 0.1;

maxR = 0.022;

numberSteps = 105;

markets = zeros(numberSteps,1);
markets(1) = k;
for i = 1:size(markets,1) - 1 
    markets(i+1) = markets(i)*(1+0.1); 
end

% ka = 1100

%%
numberSim = 1000;
distR = randsrc(numberSteps, numberSim,[-1 -.6 -.36 -.07 0 0.17 0.37 0.39 0.5 0.81 1.37 1.39 1.4 2.2 2.57 3.16 3.55 5.49 6.56 7.34 7.82 8.63 8.7 9.15 9.42 11.55 17.94 19.9 21.18 22.28 28.4 28.51;...
                                        .648 .0095 .0095 .0095 0.067 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095]);

cumR = cumsum(distR,1);

MaxD = maxdrawdown(cumR,'arithmetic');
[maxMaxD,ii] = max(MaxD);
meanD = mean(MaxD);
stdD = std(MaxD);


v = distR; % Sample data.
% Find elements where v is negative.
negativeValues = v < 0;
% Measure area and mean of the negative regions.
props = regionprops(negativeValues, v, 'MeanIntensity', 'Area');
% Extract values from structure into arrays, for convenience.
allMeanIntensities = [props.MeanIntensity];
allAreas = [props.Area];
% Get the sum of all values in each region.
integratedValues = allMeanIntensities .* allAreas;
% Find the sum that is MOST negative.
mostNegative = min(integratedValues);


figure
subplot(1,2,1)
plot(cumR)
subplot(1,2,2)
plot(cumR(:,ii))
sgtitle(['Mean Drawdown=',num2str(meanD,3),'   std=',num2str(stdD,3),'   Max Possible DD=',num2str(maxMaxD,3),'   Worst Consecutive=',num2str(mostNegative,3) ])

[Image Analyst]

I can't run it because I don't have the Communications Toolbox. Can you give me the data in a .mat file after you've created it using the Communications Toolbox function calls?

[Camilo Acevedo]

Let me know if it worked..

[Camilo Acevedo]

Hi Image Analyst, could you see the file?

Answer 2

[Mohammad Sami]

Assuming your vector is called a

a = randi([-10 10],2000,1);
ag0 = a>=0; % idx of val greater then a
id = cumsum(ag0 | circshift(ag0,1)); % create a grouping variable
% consecutive negative values are assigned to the same group
if id(1) == 0
id = id + 1; % start index at 1
end
b = accumarray(id,a,[],@sum); % sum the values for each group
[~,i] = min(b); % find the most negative sum
consective_idx = id==i; % index of the consecutive values
consective_val = a(consective_idx); % 

[Camilo Acevedo]

Thank you Mohammad. I tryed to adjust sur suggestion to my code, but I still cannot get it to work. It says that there is something wrong in the line 39 (b = accumarray(id,distR,[],@sum); % sum the values for each group).

clear variables
k = 5000;

risk = 0.007;

riskM = 0.1;

maxR = 0.022;

numberSteps = 105;

markets = zeros(numberSteps,1);
markets(1) = k;
for i = 1:size(markets,1) - 1 
    markets(i+1) = markets(i)*(1+0.1); 
end

% ka = 1100

%%
numberSim = 2;
distR = randsrc(numberSteps, numberSim,[-1 -.6 -.36 -.07 0 0.17 0.37 0.39 0.5 0.81 1.37 1.39 1.4 2.2 2.57 3.16 3.55 5.49 6.56 7.34 7.82 8.63 8.7 9.15 9.42 11.55 17.94 19.9 21.18 22.28 28.4 28.51;...
                                        .648 .0095 .0095 .0095 0.067 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095]);

cumR = cumsum(distR,1);

MaxD = maxdrawdown(cumR,'arithmetic');
[maxMaxD,ii] = max(MaxD);
meanD = mean(MaxD);
stdD = std(MaxD);


ag0 = distR>=0; % idx of val greater then a
id = cumsum(ag0 | circshift(ag0,1)); % create a grouping variable
% consecutive negative values are assigned to the same group
if id(1) == 0
id = id + 1; % start index at 1
end
b = accumarray(id,distR,[],@sum); % sum the values for each group
[~,iii] = min(b); % find the most negative sum
consective_idx = id==iii; % index of the consecutive values
consective_val = distR(consective_idx); %

Worstc = sum(consective_val, 'all')

figure
subplot(1,2,1)
plot(cumR)
subplot(1,2,2)
plot(cumR(:,ii))
sgtitle(['Mean Drawdown=',num2str(meanD,3),'   std=',num2str(stdD,3),'   Max Possible DD=',num2str(maxMaxD,3),'   Worst Consecutive=',num2str(Worstc,3) ])

[Mohammad Sami]

What is the exact error you are getting ? .

If you get the following error, you need to add a transpose to make your input into a column vector

Error using accumarray
Second input VAL must be a vector with one element for
each row in SUBS, or a scalar.

-

b = accumarray(id,distR',[],@sum); % add a transpose to distR to make it a column vector

[Camilo Acevedo]

Yes that is in deed the problem!! Thank you! Sorry but I am still learning...

[Camilo Acevedo]

@Mohammad Sami

I am trying to transpose DistR but I have not been able to do it! Any suggestions?

[Mohammad Sami]

The symbol ' will transpose the variable.

distrtranspose = distr' ;

you can also use the function transpose instead.

[Camilo Acevedo]

So simple! Thank you!

[Camilo Acevedo]

Stil not working :(

clear variables
k = 5000;

risk = 0.007;

riskM = 0.1;

maxR = 0.022;

numberSteps = 105;

markets = zeros(numberSteps,1);
markets(1) = k;
for i = 1:size(markets,1) - 1 
    markets(i+1) = markets(i)*(1+0.1); 
end

% ka = 1100

%%
numberSim = 1000;
distR = randsrc(numberSteps, numberSim,[-1 -.6 -.36 -.07 0 0.17 0.37 0.39 0.5 0.81 1.37 1.39 1.4 2.2 2.57 3.16 3.55 5.49 6.56 7.34 7.82 8.63 8.7 9.15 9.42 11.55 17.94 19.9 21.18 22.28 28.4 28.51;...
                                        .648 .0095 .0095 .0095 0.067 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095 .0095]);

cumR = cumsum(distR,1);

MaxD = maxdrawdown(cumR,'arithmetic');
[maxMaxD,ii] = max(MaxD);
meanD = mean(MaxD);
stdD = std(MaxD);

distrtranspose = transpose(distR);
ag0 = distrtranspose>=0; % idx of val greater then a
id = cumsum(ag0 | circshift(ag0,1)); % create a grouping variable
% consecutive negative values are assigned to the same group
if id(1) == 0
id = id + 1; % start index at 1
end
b = accumarray(id,distrtranspose,[],@sum); % sum the values for each group
[~,i] = min(b); % find the most negative sum
consective_idx = id==i; % index of the consecutive values
consective_val = distrtranspose(consective_idx); % 

figure
subplot(1,2,1)
plot(cumR)
subplot(1,2,2)
plot(cumR(:,ii))
sgtitle(['Mean Drawdown=',num2str(meanD,3),'   std=',num2str(stdD,3),'   Max Possible DD=',num2str(maxMaxD,3),'   Worst Consecutive=',num2str(mostNegative,3) ])

[Mohammad Sami]

What is the error you are getting.

Are you able to save the variable distR to a .mat file and upload here.

[Camilo Acevedo]

Yes. I attach the variable. The error is:

Error using accumarray

Second input VAL must be a vector with one element for each row in SUBS, or a scalar.

Error in Simulacion_sistemas_marzo2020 (line 39)

b = accumarray(id,distrtranspose,[],@sum); % sum the values for each group

[Camilo Acevedo]

Hi Mohammad, could you see the file?

[Mohammad Sami]

Hi Camilo,

I dont have the toolbox to execute your code.

Based on the function description online I think it will generate a matrix of size numberSteps x numberSims.

The code I wrote will only work for a vector of size n x 1.

In your case can I assume that you want to find the negative sequence of each column separately ?

If that is the case you can use the for loop over each column to find the sequences with the above code.

Answer 3

[Camilo Acevedo]

Well, I got some help from a friend expert in Matlab and he figured it out (based on Mohammad suggestions). So in case anyone in the future is looking for what I asked, her is the answer:

clear variables

numberSteps = 107;

numberSim = 250000;
distR = randsrc(numberSteps, numberSim,[-1 0 1.37 10.05 8.49 15.67 2.36 11.25 24.41 21.5 20.09 1.61 8.91 7.58 2.37 3.28 6.36 12.53 5.86 3.43 3.52 8.83 3.34 4.29 20.67;...
                                        .709 .084 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009 .009]);

%%
ag0 = distR>=0; 
id = cumsum(ag0 | circshift(ag0,1)); 

id = id + 1;

numCons = zeros(1,numberSim);
for a4 = 1 : numberSim
    c = accumarray(id(:,a4),distR(:,a4),[],@sum);
    [~,d] = min(c);
    numCons(a4) = sum(id(:,a4)==d);
end
maxNC = max(numCons);
meanNC = mean(numCons);
stdNC = std(numCons);