How to find mean and errorbars for scatter plot at various x values.

I have created a scatter plot from 2 matrices of the same size (501x1001) using the code

figure

scatter(Mat1(:),Mat2(:),'.');

Mat1 contains values at set intervals of 0.25 from 0 to 3, Mat2 contains many different decimal values.

I has created a scatter plot which looks like this (note, this is 4 different y matrices plotted on the same figure against the x axis), however I would like to know how I can use this to find the mean values for each of my 4 matrices and the corresponding errorbars to create more of a line graph with 4 lines and their corresponding errorbars, any help would be massively appreciated.

Thank you.

Answer 1

[William Rose]

@Charlie Milford

I don't understand the organization of Mat1 and Mat2. There are 13 x-values repeated many times among the 501,501 elements of Mat1? There are many y-values for every x value (501501/13?). I need to know how the arrays are structured, in order to find the mean and SD at each x-value. Can you represent the data as

x=0:.25:3;
y1=randn(501,13);
y2=0.5+randn(501,13)*1.5;
y3=-0.5+randn(501,13)*2;
%Next findmean and SD ateach x-value
y1m=mean(y1);
y1sd=std(y1);
y2m=mean(y2);
y2sd=std(y2);
y3m=mean(y3);
y3sd=std(y3);

%plot results
figure;
errorbar(x,y1m,y1sd,'r')
hold on;
errorbar(x,y2m,y2sd,'g')
errorbar(x,y3m,y3sd,'b')
legend('y1','y2','y3')

That would be much more compact for the x-values, and easier to handle.

Good luck!

[Charlie Milford]

My 'x' values in 'Mat1' are also a matrix of the same size of all my other 'y' or Mat2 matrices. But all values in the x matrix are from 0 to 3 in 0.25 intervals.

I think what I have done here has worked:

meandif=[0:0.25:3];

for i=1:13

a=find(Stanprog_region==meandif(i));

means=nanmean(PDdifference10(a));

stds=nanstd(PDdifference10(a));

meandif10(i)=means;

stddif10(i)=stds;

end

%Where Stanprog_region=Mat1 and PDdifference10=Mat2

Thanks for your help.

Answer 2

[dpb]

The specific answer is

M1=mean(Mat2,'all');
S1=std(Mat2,[],'all');

general answer is to not create multiple sequentially-named variables but use either a 3D array where each is by plane or a cell array where each cell holds the ith 2D array. Then you can write generic code that iterates over the array.

The x value for plotting would then be just be the single x vector.

[Charlie Milford]

Hi, thanks for your answer, I tried a for loop for each array and this seem to work, do you think this is correct?

meandif=[0:0.25:3];
for i=1:13
    a=find(Stanprog_region==meandif(i));
    means=nanmean(PDdifference10(a));
    stds=nanstd(PDdifference10(a));
    meandif10(i)=means;
    stddif10(i)=stds;
end

[dpb]

Well, really couldn't say without knowing how the data were stored/content of the variables and what actually was intended to compute, but if it looks about right, it probably is.

You can always create small-enough arrays or use subsets of the given arrays that are small enough you can see precisely what data you have and manually check get the answers expected.

[Charlie Milford]

This is what I produced which seems to be the trend I was looking for. Thanks a lot for your help!