How to fit a particular analytical function while optimising its parameters to a histogram for different data sets in a loop in Matlab?

Hi,

I am trying to fit a particular analytical function to histogram of a dataset. The parameters of the function is being optimised using lsqcurvefit command of Matlab. When I am trying for a single data set, its giving right curve. But when I am trying to put it in a loop to fit several datasets such that each time it will take one dataset and fit the function with optimized parameters to its histogram and then go to the next dataset and do the same job, it is giving wrong curve.

I am suspecting that somehow it is taking garbage values while plotting the function.

please help me out with this. I am attaching input 3 datasets .

thank you.

d=dir('latlong*.csv');   						
n=length(d);							
data=cell(n,1);

for i = 0:2
   for j=0:0
       data{j*70 + i+1}=csvread(d(j*70 + i+1).name,1,0,[1,3,8830,3]);
       a = data{j*70 + i+1};
       b(:,1) = a(:,2);

        k3=0;
        for l=1:8831		
               if b(l,1)>=0.001 
                   k3=k3+1;
               end
        end
        z=zeros(k3,1);
        k3=0;
        for l=1:8831		
               if b(l,1)>=0.001 
                   k3=k3+1;
                   z(k3,1)=b(l,1);
               end
        end
        
        figure(i+1);
        hold on
        h=histogram(z,1000,'Normalization','pdf');
        no=h.NumBins;
        counts=zeros(no,1);
        counts(:,1)=h.Values;
        edge=zeros(no+1,1);
        edge(:,1)=h.BinEdges;
        edge=edge(2:end,1);

        Fun3=@(q,edge) ((1./(edge.*q(2).*sqrt(2*pi))).*exp(-((log(edge)-q(1)).^2)./(2*q(2)^2)));
        q0=[2,2];
        q(i+1,:) = lsqcurvefit(Fun3,q0,edge,counts);
        plot(edge,Fun3(q,edge));
       
        hold off
        xlabel('Observation')
        ylabel('Probability Density')
        legend('histogram','log')
    end
end

Answer 1

[Brian]

Hello Joya,

I refactored your code to get a better idea of what it was doing, and there seem to be two issues. The main issue is that on line 39 you call "Fun3" with "q" instead of "q(i+1,:)". The second issue is that "csvread" is prepending a zero to your data. The "csvread" function is not recommended and it is recommended to use "readmatrix" or "readtable" instead.

d=dir('latlong*.csv');   						
n=length(d);							
data=cell(n,1);
q=zeros(n,2);

for i = 1:n
    data{i}=readtable(d(i).name);
    z=data{i}.precipitation(data{i}.precipitation >= 0.001);
    figure(i);
    hold on
    h=histogram(z,1000,'Normalization','pdf');
    no=h.NumBins;
    counts=h.Values';
    edge=h.BinEdges(2:end)';
    Fun3=@(q,edge) ((1./(edge.*q(2).*sqrt(2*pi))).*exp(-((log(edge)-q(1)).^2)./(2*q(2)^2)));
    q0=[2,2];
    q(i,:)=lsqcurvefit(Fun3,q0,edge,counts);
    plot(edge,Fun3(q(i,:),edge));
   
    hold off
    xlabel('Observation')
    ylabel('Probability Density')
    legend('histogram','log')
end

Hope this helps, let me know if I misunderstood anything.

[JOYA GHOSH DASTIDER]

Thank you for your reply and really sorry for being late in acknowledging .

I also figured out the problem of not using "q(i+1,:)".

Thanks again for suggesting a compact form and use of "readtable" instead of "csvread".