How do I display different boxplot groups on the same figure in MATLAB?

I would like to generate boxplots on the same figure for two different groups, observed and simulated temperatures for January to December with the x-axis being months. The two different groups need to be represented in different colors. The spacing between the boxplots also need to be user-definable.

Answer 1

[Doug Hull]

There is no direct way of displaying boxplots for two different groups (in this example, observed and simulated temperatures) on the same figure.

As a workaround, boxplots can be generated at defined positions for one group first. HOLD ON allows the boxplots for the second group to display on the same figure.

In the following example, boxplots of the second group are 0.2 units away (user defined) from their corresponding blue boxplots.

% {Display boxplots for two different groups, 
% observed and simulated temperatures from January to December on the same figure}%  
close all  
clc 
data = textread('bp.txt');  % Read user’s data file   
%{Assign the observed temperature to be Month_O and the simulated 
% temperature to be Month_S}% 
Jan_O = data(:, 1); 
Jan_S = data(:, 2); 
Feb_O = data(:, 3); 
Feb_S = data(:, 4); 
Mar_O = data(:, 5); 
Mar_S = data(:, 6); 
Apr_O = data(:, 7); 
Apr_S = data(:, 8); 
May_O = data(:, 9); 
May_S = data(:,10); 
Jun_O = data(:,11); 
Jun_S = data(:,12);
Jul_O = data(:,13); 
Jul_S = data(:,14); 
Aug_O = data(:,15); 
Aug_S = data(:,16); 
Sep_O = data(:,17); 
Sep_S = data(:,18); 
Oct_O = data(:,19); 
Oct_S = data(:,20); 
Nov_O = data(:,21); 
Nov_S = data(:,22); 
Dec_O = data(:,23); 
Dec_S = data(:,24);   

f=figure;   
% Boxplot for the observed temperature from January to December 

Temp_O = [Jan_O, Feb_O, Mar_O, Apr_O, May_O, Jun_O, Jul_O, Aug_O, Sep_O, Oct_O, Nov_O, Dec_O]; 

position_O = 1:1:12;  
% Define position for 12 Month_O boxplots  

box_O = boxplot(Temp_O,'colors','b','positions',position_O,'width',0.18); 
set(gca,'XTickLabel',{' '})  % Erase xlabels   
hold on  % Keep the Month_O boxplots on figure overlap the Month_S boxplots   
% Boxplot for the simulated temperature from January to December 

Temp_S = [Jan_S, Feb_S, Mar_S, Apr_S, May_S, Jun_S, Jul_S, Aug_S, Sep_S, Oct_S, Nov_S, Dec_S]; 
position_S = 1.3:1:12.3;  % Define position for 12 Month_S boxplots  
box_S = boxplot(Temp_S,'colors','r','positions',position_S,'width',0.18);   

hold off   % Insert texts and labels 
ylabel('Temperature') 
text('Position',[1.1,0],'String','January') 
text('Position',[2.1,0],'String','February') 
text('Position',[3.1,0],'String','March') 
text('Position',[4.1,0],'String','April') 
text('Position',[5.1,0],'String','May') 
text('Position',[6.1,0],'String','June') 
text('Position',[7.1,0],'String','July') 
text('Position',[8.1,0],'String','August') 
text('Position',[9.1,0],'String','September') 
text('Position',[10.1,0],'String','October') 
text('Position',[11.1,0],'String','November') 
text('Position',[12.1,0],'String','December') 

set(gca,'XTickLabel',{''});   % To hide outliers 
out_O = box_O(end,~isnan(box_O(end,:)));  
delete(out_O)  
out_S = box_S(end,~isnan(box_S(end,:)));  
delete(out_S)

[Martha]

This example just saved my life.

[Mark Cejas]

This is a great example! thanks

[Julia]

I copied this code and substituted my data, but the x labels aren't showing up. Any advice? Thanks.

[Kevin Heimbach]

amazing example, needed this exactly

[Sayantan Sahu]

What should I do to get the 5th and 95th percentile ?

[Denise Napolitano]

Excellent code, thank you! Do you have any advice on getting the object handles to modify box properties? I usually use a = get(get(gca,'children'),'children'); but 'a' is interpreted as a 2x1 cell array (21x1 lines, since there are 3 sets of stacked data):

a =

2×1 cell array

    [21×1 Line]
    [21×1 Line]

[Ellyn Gray]

Thank you so much!!! I spent hours trying to do this with boxplot and boxplot2 and your solution worked in 10 mins. Exactly what I wanted and your example was super easy to follow.

Answer 2

[Tom Lane]

The boxplot function has more options than you can shake a stick at. Try this:

data = rand(20,24)
month = repmat({'jan' 'feb' 'mar' 'apr' 'may' 'jun' 'jul' 'aug' 'sep' 'oct' 'nov' 'dec'},1,2);
simobs = [repmat({'sim'},1,12),repmat({'obs'},1,12)];
boxplot(data,{month,simobs},'colors',repmat('rb',1,12),'factorgap',[5 2],'labelverbosity','minor');

[Ville-Pekka Seppä]

Thanks, this is quite helpful. In my case however the data consists of 9 categories each divided into 4 subcategories which do not contain equal amount of values.

I would like to plot this divided into nice groups (each having 4 boxes) and be able to choose the order of the groups. Especially choosing the order of the groups is giving me trouble as the Matlab help is rather vague on that for multiple grouping variables.

Any help appreciated!

[Tom Lane]

Consider making ordinal variables to specify the order. For example:

load carsmall
boxplot(MPG,{Origin Model_Year})  % default ordering

org = ordinal(Origin,[],{'France' 'Germany' 'Sweden' 'Italy' 'Japan' 'USA'});
boxplot(MPG,{org Model_Year})     % European countries grouped together

yr = ordinal(Model_Year,[],[82 76 70]);
boxplot(MPG,{org yr})             % years running backward

[Ville-Pekka Seppä]

Ordinal variables were the key to this. Thanks!

[Berk]

Tom, is it possible to apply your code in your initial reply to situations where the groups do not have the same number of observations? For example, in your example each Month/Sim or Month/Obs couple has 20 observations. Would it be possible to make this more flexible so that Obs, for example, has more observations than Sim?

[Tom Lane]

The example that begins "load carsmall" has different numbers of observations for each value or Origin. So arranging your data in a vector with a corresponding grouping value is one way to do what you want.

In the example with "simobs" I set up a matrix, so it naturally has the same number of values in each column. It would be possible to pad each column with NaN values if there were not an equal number of numeric values for each column.

[Julia]

I didn't get how to set up my data for this code. I tried having 24 columns, with the first 12 being the first set and the second 12 being the second set. The data were lumped together instead of being displayed side-by-side.

[Tom Lane]

My May 2012 example shows how to do this. You'll need to explain what you want, what you did, and what you got if you want more advice.

[Algis]

This implementation seems to be wrong, because for each boxplot it takes not 20 elments, but somehow 40. Can you suggest on how to fix it?

[Hillaryfor2016]

I couldn't get this method to work for my code. I instead used a function 'boxplot2' and it worked a dream. See http://uk.mathworks.com/matlabcentral/answers/176766-multiple-boxplots-for-a-single-plot

[Mohammad Naser]

Thanks

Answer 3

[Mohammad Abu Zafer Siddik]

Hi Tom Lane,

Thank you for your code, i usee your code in the folowing way.

boxplot(num,'colors',repmat('gbrm',1,13),'PlotStyle','compact','symbol','+');

However, i need to give RGB such as "[25 150 81]./255" for green and so on for other colors. Because, the default green is not clear when we print black and white. Would you please tell me a way that how can i use RGB for different darker colors?

Thanks

Mohammad

[Ellyn Gray]

This is the most useful thing I've found for finding colors. It can be a bit of a guessing game testing colors without it.

https://www.mathworks.com/matlabcentral/fileexchange/24497-rgb-triple-of-color-name-version-2