From a statistical perspective, the boxplot is correct (plots medians and percentiles) but for your application, this should work:
data = bsxfun(@times, rand(5,3), [50 150 100]); % Create Data
dmean = mean(data); % Mean
dci = std(data)*tinv(0.975,size(data,1)-1); % Confidence Intervals
xt = [1:3]; % X-Ticks
xtd = repmat(xt, size(data,1), 1); % X-Ticks For Data
sb = [xt'-ones(size(data,2),1)*0.1, xt'+ones(size(data,2),1)*0.1]; % Short Bar X
lb = [xt'-ones(size(data,2),1)*0.2, xt'+ones(size(data,2),1)*0.2]; % Long Bar X
figure(1)
plot(xt, data, '+')
hold on
for k1 = 1:size(data,2)
plot(lb(k1,:), [1 1]*dmean(k1), sb(k1,:), [1 1]*(dmean(k1)-dci(k1)), sb(k1,:), [1 1]*(dmean(k1)+dci(k1)), '-k')
end
hold off
set(gca, 'XTick', xt, 'XTickLabel', {'Left','Middle','Right'})
xlabel('Group')
ylabel('Velocity (Furlongs/Fortnight)')
I plotted ±95% confidence intervals, but if you want to plot something else, just change the ‘dci’ assignment. The rest of the code will adapt automatically.
