I'm having a problem averaging multiple curves using interp1

Hi everyone.

I have multiple polarisation curves that I want to display the averge of. I tried using linspace to create a base vector and interpolating using interp1. Unfortunately that hasn't properly worked for me and I was hoping someone might be able to help.

Thank you in advance!

Answer 1

[Chris]

1. You take the mean and max of the U values; I believe you want the I values instead.
2. You have plenty of data points, so the default linear interpolation will follow the trend better.
3. Some data at the end will have to be excluded from the mean curve. You could use the 'omitnan' flag, but that will cause a discontinuity in the curve.

%  Mittelung mehrerer Messungen
clearvars
[filenames, pathname] = uigetfile('MultiSelect', 'on', '*.*');
fullname = fullfile(pathname,filenames);
clear savename

for z = 1:length(fullname)
    load (fullname{1,z})
    loadDaten{1,z} = Daten;
end

Daten = loadDaten;
IVC_mean = cell (3,length(fullname));
var = zeros(1,length(fullname));
Names = string(var);
Imax = zeros (1,length(fullname));
Umax = zeros (1,length(fullname));
Umin = zeros (1,length(fullname));

for z = 1:length(fullname)
    
    % Messdaten
    Ewe = Daten{1,z}(:,7);
    I = Daten{1,z}(:,8).*1000;
    Ismooth = smoothdata(I,'sgolay');

    % Details der Messung
    savename{1,z} = extractBefore(filenames{1,z},".");
    Names(z) = savename {1,z};

    % sortieren
    IVC_mean{1,z} = Ewe;
    IVC_mean{2,z} = abs(Ismooth);
    IVC_mean{3,z} = extractAfter(strrep(savename{1,z},'_',' '),' ');
%     % outlier
%     pp = isoutlier(IVC_mean{2,z});
%     ind = find(pp);
%     IVC_mean{4,z} = ind;
%     IVC_mean{5,z} = IVC_mean{2,z};
%     IVC_mean{5,z}(ind) = NaN;

    % einzeln plot
    h = scatter(IVC_mean{2,z},IVC_mean{1,z});
    xlabel(['I']);ylabel(['U']);
    hold on

    % Grenzen für xq
%     IVC_mean{6,z} = min(IVC_mean{1,z});
%     IVC_mean{7,z} = max(IVC_mean{1,z});
    IVC_mean{6,z} = min(IVC_mean{2,z});
    IVC_mean{7,z} = max(IVC_mean{2,z});
    Umin(z) = IVC_mean{6,z};
    Umax(z) = IVC_mean{7,z};
end
%
% Interpolation
Umin = min(Umin);
Umax = max(Umax);
% vorgegebener Bezugsvektor 
UC = linspace(Umin,Umax,10000);

for z = 1:length(fullname)
%     IVC_mean{8,z} = interp1(IVC_mean{2,z},IVC_mean{1,z},UC,'spline');
    IVC_mean{8,z} = interp1(IVC_mean{2,z},IVC_mean{1,z},UC,'linear');

    h2 = plot(UC,IVC_mean{8,z},'k--','LineWidth',2);
    hold on
end

mfit = mean(cat(1,IVC_mean{8,:}));
plot(UC, mfit,'m','LineWidth',2);

[Philip Krämer]

I've tried usign the I values before but when I do it says the sample points must be unique and quite a lot of them aren't. Technically i measured the current over a linear potential sweep, so I should be fine using the U values I think. Displaying U over I is just the conventional thing to do...

[Walter Roberson]

Some data at the end will have to be excluded from the mean curve

They are using 'spline' which does interpolation by default, so none of the results will be nan.

Perhaps it is "good enough" to interpolate at the tails, perhaps not. But if not then instead of taking the min() of the min() and the max of the max, Umin = max(Umin) and Umax = min(Umax) would create an interval that included only the portion common to all three curves, which would avoid any need for interpolation.

[Chris]

@Philip Krämer As Walter pointed out, the script you uploaded used the range of IVC_mean{1,:}, which had been assigned the values of Ewe. The script then fits U values along that reduced range ([0.3, 1.3]). The splines go wild outside of the interpolated range. You can fit U along the range of I just by changing three lines:

IVC_mean{6,z} = min(IVC_mean{2,z});
IVC_mean{7,z} = max(IVC_mean{2,z});

Then:

h2 = scatter(UC, IVC_mean{8,z});

Since the splines still go wild on the shorter fits, you can zoom in with the axes to see your range of interest (or plot along a reduced range of I/UC)

xlim([0.1 1.6])

[Philip Krämer]

Thanks for your help!

[Chris]

Bitte sehr!

Answer 2

[Walter Roberson]

    h = scatter(IVC_mean{2,z},IVC_mean{1,z});

So {1} is used as y values and {2} is used as x values.

    IVC_mean{6,z} = min(IVC_mean{1,z});
    IVC_mean{7,z} = max(IVC_mean{1,z});

min and max of the y values.

    Umin = min(Umin);
    Umax = max(Umax);
    UC = linspace(Umin,Umax,10000);

smallest y and greatest y

IVC_mean{8,z} = interp1(IVC_mean{2,z},IVC_mean{1,z},UC,'spline');

you pass in known x values and corresponding known y values and you query based on UC, which is based on y values, not on x values.