{"id":3798,"date":"2012-08-17T06:30:40","date_gmt":"2012-08-17T11:30:40","guid":{"rendered":"https:\/\/blogs.mathworks.com\/pick\/?p=3798"},"modified":"2012-08-17T06:30:40","modified_gmt":"2012-08-17T11:30:40","slug":"shaded-error-bars","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/pick\/2012\/08\/17\/shaded-error-bars\/","title":{"rendered":"Shaded Error Bars"},"content":{"rendered":"
\r\n

Jiro<\/a>'s pick this week is shadedErrorBar<\/tt><\/a> by Rob Campbell<\/a>.\r\n <\/p>\r\n

This one was suggested<\/a> to us by Oliver<\/a>, but I'm pretty certain I would have picked it if I had come across this entry.\r\n <\/p>\r\n

In fact, we used to (try to) make similar plots back in my graduate school days. We had some time series data from multiple\r\n subjects, and we wanted a way to see trends and variability. It took a little while for me to arrive at this way of visualizing\r\n variability. Initially, I would simply plot the line series and visually inspect how close or separated the lines were. We\r\n already had quantitative measures, but I also wanted the visual representation.\r\n <\/p>

% From one of Rob's examples<\/span>\r\nx=(1:80)-40;\r\ny=ones(15,1)*x; y=y+0.06*y.^2+randn(size(y))*10;\r\n\r\nplot(x, y);<\/pre> 
Next, I simply created lines to represent the mean and multiple standard deviations. This was better, but it was still a bit\r\n      crowded visually.\r\n   <\/p>
mean_y = mean(y);\r\nstd_y = std(y);\r\n\r\nfigure;hold on<\/span>;\r\n\r\nH1 = plot(x, mean_y, 'Color'<\/span>, 'k'<\/span>, 'LineWidth'<\/span>, 2);\r\nH2 = plot(x, [mean_y - 2*std_y; mean_y + 2*std_y], 'Color'<\/span>, 'r'<\/span>);\r\nH3 = plot(x, [mean_y - std_y; mean_y + std_y], 'Color'<\/span>, 'm'<\/span>);\r\nH4 = plot(x, [mean_y - 0.5*std_y  ; mean_y +   0.5*std_y], 'Color'<\/span>, 'b'<\/span>);\r\n\r\nlegend([H1, H2(1), H3(1), H4(1)], ...<\/span>\r\n    '\\mu'<\/span>, '2\\sigma'<\/span>, '\\sigma'<\/span>, '0.5\\sigma'<\/span>, ...<\/span>\r\n    'Location'<\/span>, 'Northwest'<\/span>);<\/pre> 
Finally, we decided on a shaded representation of bounds. I don't remember how complicated the code was, but with Rob's shadedErrorBar<\/tt>, I can do this very easily!\r\n   <\/p>
figure;hold on<\/span>;\r\n\r\nH(1) = shadedErrorBar(x, y, {@mean, @(x) 2*std(x)  }, '-r'<\/span>, 0);\r\nH(2) = shadedErrorBar(x, y, {@mean, @(x) 1*std(x)  }, '-m'<\/span>, 0);\r\nH(3) = shadedErrorBar(x, y, {@mean, @(x) 0.5*std(x)}, {'-b'<\/span>, 'LineWidth'<\/span>, 2}, 0);\r\n\r\nlegend([H(3).mainLine, H.patch], ...<\/span>\r\n    '\\mu'<\/span>, '2\\sigma'<\/span>, '\\sigma'<\/span>, '0.5\\sigma'<\/span>, ...<\/span>\r\n    'Location'<\/span>, 'Northwest'<\/span>);<\/pre> 
The code is well-written, with help text, examples, error-checking, and lots of comments – everything that I look for in a\r\n      good MATLAB code! Thanks for your entry, Rob, and thanks Oliver for the suggestion!\r\n   <\/p>\r\n   
Comments<\/b><\/p>\r\n   
Let us know what you think here<\/a> or leave a comment<\/a> for Rob.\r\n   <\/p>