Legible error on multiple points for same x-value

I have data points which look like this when plotted:

they represent multiple trials for the same conditions.

I want to add some sort of statistical error such as error bars or confidence bounds. However error bars are not really legible since they overlap at this scale and type.

So what are some alternatives?

This is some sample data:

    Force:     Height:
    254.3150   15.0000
    250.6990   15.0000
    186.1220   15.0000
    114.3150   10.0000
    147.1020   10.0000
    115.8110   10.0000
    114.6890   10.0000
     95.9890   10.0000
     98.1080   10.0000
     75.9160    7.5000
     73.9220    7.5000
     58.5950    5.0000
     65.4450    5.0000
     57.5920    5.0000
     75.7930    5.0000
     51.2340    5.0000
     14.9620    2.5000
     14.7120    2.5000
     17.4480    2.5000
      1.3730         0
      1.2480         0

Answer 1

[Thorsten]

Well, you have good data, that's what your overlapping errorbars show. And that's why you show them. If all your data would be like your measurements for 0, 2.5 and 7.5 you could write " Error bars were on the scale of the symbols and are not shown." But in your case you have some visible error bars for the other measurements, so I think its just fine to show them.

You could do it like this

x = X(:,2); y = X(:,1);

y_groups = {};
x_unique = unique(x(:))';
for x_u= x_unique
  y_groups{end + 1} = y(x_u == x); 
end

my = cellfun(@(x) mean(x), y_groups);
sy = cellfun(@(x) std(x), y_groups);
plot(x, y, '.r')
hold on
errorbar(x_unique, my, sy, 'ks')