How to plot mean scores of dataset with different number of variables

I have a matrix of 2922 samples with scores for 23 components (2922x23) double.

I want to plot the mean value of the first component for data points 1-320, 321-655, 656-1001, 1002-1338, 1339-1684, 1685-2027, 2028-2348, 2349,2581 and 2582-2922 including an error bar of +/-1 standard deviation for each group. The components are not numerical values.

What would be the simplest way to do this for each component? The number of data points within each set will never be the same. Ideally I would like to group the 9 "sets" into 2 groups of 4 and 5 which have different coloured plots.

Answer 1

[the cyclist]

% Make up some pretend data
data = rand(2922,23);

% 1-320, 321-655, 656-1001, 1002-1338, 1339-1684, 1685-2027, 2028-2348, 2349,2581 and 2582-2922
edge = [0 320 655 1001 1338 1684 2027 2348 2581 2922];

% Find the number of sections
numberSections = numel(edge)-1;

% Preallocate memory for mean and std dev
sectionMean = nan(numberSections,1);
sectionStd  = nan(numberSections,1);

% Calculate the stats for each section
for ne = 1:numberSections
    indexToSection = edge(ne)+1:edge(ne+1);
    sectionMean(ne) = mean(data(indexToSection,1));
    sectionStd(ne)  = std(data(indexToSection,1));
end

% Plot
figure
errorbar(sectionMean,sectionStd)

[Kirsty Milligan]

Thanks you for this!

I should have added that I would like to plot the values for each column individually, so a plot of the values as described above (e.g. rows still split into 1-320... etc) but for column 1 and column 2 and so on... separately.

[the cyclist]

Just wrap a for loop around it.

[Kirsty Milligan]

Thanks! One final question, if I wanted to calculate the standard error as opposed to the standard deviation how could I incorporate this?

[the cyclist]

Making an assumption of normality, the standard error (of the mean) is the standard deviation divided by the square root of the number of samples. So I think the following should be correct.

sectionStdErr(ne) = sectionStd(ne)/sqrt(numel(data(indexToSection,1)));

[Kirsty Milligan]

This gives no error but calculates all sectionStdErr(ne) valules as NaN for some reason

[Kirsty Milligan]

I think the problem there was a sight typo with the Std, however this still gives me values of 0 and an undefined function error on "sectionStd".

Using just

sectionStdErr(ne) = std(ne)/sqrt(numel(data(indexToSection,1)));

gives values of zero.

[the cyclist]

You definitely do not want just

std(ne)

because ne is a scalar, and std(ne) is going to be zero. It is in no way related to the standard deviation of your data.

This addition to my code above works for me:

% Make up some pretend data
data = rand(2922,23);
% 1-320, 321-655, 656-1001, 1002-1338, 1339-1684, 1685-2027, 2028-2348, 2349,2581 and 2582-2922
edge = [0 320 655 1001 1338 1684 2027 2348 2581 2922];
% Find the number of sections
numberSections = numel(edge)-1;
% Preallocate memory for mean and std dev
sectionMean = nan(numberSections,1);
sectionStd  = nan(numberSections,1);
sectionStdErr  = nan(numberSections,1);
% Calculate the stats for each section
for ne = 1:numberSections
    indexToSection = edge(ne)+1:edge(ne+1);
    sectionMean(ne) = mean(data(indexToSection,1));
    sectionStd(ne)  = std(data(indexToSection,1));
    sectionStdErr(ne) = sectionStd(ne)/sqrt(numel(data(indexToSection,1)));
end
% Plot
figure
errorbar(sectionMean,sectionStdErr)

The modification is that I added the calculation of the standard error, and plotted it instead of the standard deviation.

[Kirsty Milligan]

Thank you so much! Works perfectly.