Bar plot with two y axes

Question

I have the following code for a plot with 2 y-axes in MATLAB. I am glad that the 2-axes feature works, however, I would like to avoid the overlapping of the bars. Also, the categories on the right-hand axis should have different colors, not only yellow, yet it should be somehow clear that they are plotted on the right-hand axis and not the left one. How can this be done?

EONMW = [100 399 500];
RWEMW = [200 996 120];
GermanByEON = [0.2 0.4 0.5];
GermanByRWE = [0.1 0.5 0.9];
EONGermanPortfolio = [0.7 0.2 0.1];
RWEGermanPortfolio = [0.8 0.3 0.6];
years = [2010 2012 2014];
% Plot
values1 = [EONMW; RWEMW]';
values2 = [GermanByEON; GermanByRWE; EONGermanPortfolio; RWEGermanPortfolio]';
years1 = [years; years]';
years2 = [years; years; years; years]';
figure;
bar(years1,values1);
ylabel('Utilities generation portfolio in MW')  
yyaxis right
bar(years2,values2);
legend('EON German portfolio in MW', 'RWE German portfolio in MW',...
    'Percentage of German portfolio by EON', 'Percentage of German portfolio by RWE',...
    'EON"s percentage of generation in Germany', 'RWE"s percentage of generation in Germany')
legend('Location','northwest')
ylabel('Utilities generation portfolio as percentages')

The edit makes the question clearer, but the fact you've got 3 different metrics on a single plot is pretty unclear from a (personal) data visualisation perspective... consider just using `subplot` instead, you can split your data into stacked plots so they line up on the x axis but the purpose and content of each plot is clearer. It will also save you getting bogged down in workarounds for the above, which I'm not sure will be easy for `bar`. The reason all `yyaxis right` bars are the same colour is so you know which axis they lie on - how would you know otherwise? — Wolfie, Jan 07 '19 at 09:59
Related from same OP: https://stackoverflow.com/questions/54071530/bar-plot-switch-colors-in-matlab — SecretAgentMan, Jan 07 '19 at 22:36
I would not recommend plotting the data like this (even with the [suggestion](https://stackoverflow.com/a/54072510/2627163) of @Dev-iL). I can't see any advantage in plotting these bar plots with 2 y-axes, instead of using 2 vertical subplots, with aligned years. Why plotting the bars together if they can't be compared anyway? — EBH, Jan 08 '19 at 07:02

score 7 · Answer 1 · answered Jan 07 '19 at 10:25

I agree that it is somewhat difficult to read this sort of plot, but perhaps there's a way to improve the situation a little.

The main thing I changed was adding invisible bars (using NaN) so that the color order remains intact:

function q54071610
EONMW = [100 399 500];
RWEMW = [200 996 120];
GermanByEON = [0.2 0.4 0.5];
GermanByRWE = [0.1 0.5 0.9];
EONGermanPortfolio = [0.7 0.2 0.1];
RWEGermanPortfolio = [0.8 0.3 0.6];
years = [2010 2012 2014];
% Plot
values1 = [[EONMW; RWEMW].' NaN(3,4)];
values2 = [NaN(3,2) [GermanByEON; GermanByRWE; EONGermanPortfolio; RWEGermanPortfolio].'];
years1 = repmat(years.', [1,6]);
figure;
bar(years1,values1, 'EdgeColor', 'b', 'LineWidth', 2);
ylabel('Utilities generation portfolio in MW')  
yyaxis right
hB = bar(years1,values2, 'EdgeColor', 'r', 'LineWidth', 2);
c = lines(6);
for ind1 = 1:6
  hB(ind1).FaceColor = c(ind1, :);
end
legend('EON German portfolio in MW', 'RWE German portfolio in MW',...
    'Percentage of German portfolio by EON', 'Percentage of German portfolio by RWE',...
    'EON"s percentage of generation in Germany', 'RWE"s percentage of generation in Germany')
legend('Location','northwest')
ylabel('Utilities generation portfolio as percentages')

Which results in:

Now you should explain that red-framed bars belong to the right axis whereas blue-framed bars belong to the left.

This is nice, thank you! Only problem: in the legend the entries show up with only blue frames... — LenaH, Jan 07 '19 at 11:25
Yes, I noticed this. Not sure what can be done (editing the legend entries should be possible but likely difficult). What you can do is maybe plot another bar plot, in advance, with all `NaN` y-values and then show the legend (which will have small black frames around entries, as in your image). Just add `bar(years1,NaN*values1); hold on;` after `figure;` and see if the result works for you. — Dev-iL, Jan 07 '19 at 12:00

EBH · Accepted Answer · 2019-01-08T08:16:37.913

I'm not sure what exactly these bars mean, and so I may have missed the point of the figure (which could be the main problem here). However, I find this way of presentation not pleasing and misleading, as it takes a lot of effort from the reader to understand which value belongs to which bar, and what is comparable and what's not.

What I suggest here, is not a direct answer to the technical problem (which you have already got from @Dev-iL), but a different solution for the more basic problem - how to visualize these data? I believe that if I'll understand what the numbers represent (percentage from what?) and what you want to emphasize with this plot, I can find a better solution.

First, the code:

EONMW = [100 399 500];
RWEMW = [200 996 120];
GermanByEON = [0.2 0.4 0.5];
GermanByRWE = [0.1 0.5 0.9];
EONGermanPortfolio = [0.7 0.2 0.1];
RWEGermanPortfolio = [0.8 0.3 0.6];
years = [2010 2012 2014].';
values1 = [EONMW; RWEMW].';
values2 = [GermanByEON; GermanByRWE; EONGermanPortfolio; RWEGermanPortfolio].'*100;

% Plot
colMap = mat2cell(lines(2),[1 1],3); % Choose your favorite colors
figure(2);
% upper plot:
subplot 211
b = bar(years,values1);
set(b,{'FaceColor'},colMap)
xticklabels({}) % remove the years labels, the bottom axes will show them
ylabel('Utilities generation portfolio in MW')
legend('EON German', 'RWE German',...
    'Location','northwest')

% bottom plot
subplot 212
b = bar(years,values2);
set(b,{'FaceColor'},repmat(colMap,2,1)) % matching the colors by topic
set(b,{'FaceAlpha'},{1;1;0.6;0.6}) % distiguish between related mesures
xlabel('Year')
ylabel('Utilities generation portfolio (%)')
legend('German portfolio by EON', 'German portfolio by RWE',...
    'EON''s generation in Germany', 'RWE''s generation in Germany',...
    'Location','north')

The result:

The major things I changed:

Split the bars by the units of the y-axis, but align them by the x-axis
Match the colors of related bars between the plots
Shorten legends and labels

Good luck!

score 0 · Answer 3 · answered Jan 07 '19 at 12:00

It is best if you define bar(x,y) as b=bar(x,y) then control the options of b one of such is b.FaceColor. Take a look at the following code

EONMW = [100 399 500];
RWEMW = [200 996 120];
GermanByEON = [0.2 0.4 0.5];
GermanByRWE = [0.1 0.5 0.9];
EONGermanPortfolio = [0.7 0.2 0.1];
RWEGermanPortfolio = [0.8 0.3 0.6];
years = [2010 2012 2014];
x=0.2;
% Plot
values1 = [EONMW; RWEMW]';
values2 = [GermanByEON; GermanByRWE; EONGermanPortfolio; RWEGermanPortfolio]';
years1 = [years; years]';
years2 = [years; years; years; years]';
figure;
b1=bar(years1,values1,x*0.66);
ylabel('Utilities generation portfolio in MW')  
yyaxis right
b2=bar(years2,values2,x);
%%%%%%%%%%%%
%%%%%%%%%%%%
%% Defining colors
b1(1).FaceColor=[1 0 0];
b1(2).FaceColor=[0 1 0];
b2(1).FaceColor=[0 0 1];
b2(2).FaceColor=[1 1 0];
b2(3).FaceColor=[0 1 1];
b2(4).FaceColor=[1 0 1];
%%%%%%%%%%%%
%%%%%%%%%%%%
%%
legend('EON German portfolio in MW', 'RWE German portfolio in MW',...
    'Percentage of German portfolio by EON', 'Percentage of German portfolio by RWE',...
    'EON"s percentage of generation in Germany', 'RWE"s percentage of generation in Germany')
legend('Location','northwest')
ylabel('Utilities generation portfolio as percentages')

which is your code plus the following block

%%%%%%%%%%%%
%%%%%%%%%%%%
%% Defining colors
b1(1).FaceColor=[1 0 0];
b1(2).FaceColor=[0 1 0];
b2(1).FaceColor=[0 0 1];
b2(2).FaceColor=[1 1 0];
b2(3).FaceColor=[0 1 1];
b2(4).FaceColor=[1 0 1];
%%%%%%%%%%%%
%%%%%%%%%%%%
%%

and the result is as follows

also a trigger option x has been added to control bar width.

In the code above, after defining b=bar(x,y), b becomes a vector each component being the corresponding **bar control panel**. For example if b=bar(x,y) where x and y are 2*20 vectors, then b maintains a 2*1 vector with b(1) relating to bar(x(1,:),y(1,:)) and b(2) relating to bar(x(2,:),y(2,:)) — Mostafa Ayaz, Jan 07 '19 at 17:04
@mostafa I believe he meant to ask how does the viewer understand if they should read the y value from the left or the right axis? — Dev-iL, Jan 07 '19 at 17:07
Do you mean when printed with a black-ink on paper(for which colors become indistinct)? Otherwise, I think the colors and the legend are self explaining.... — Mostafa Ayaz, Jan 07 '19 at 18:17
@MostafaAyaz Even before printing... Since we have two y-axes which mean different things (it's not a matter of just converting units, the _meaning_ of the axes is completely different). Now, each bar should be read **according to the correct y-axis**, so the question is - how does your suggestion tell the reader if they should look at the right axis or the left axis? I hope it's clear now. — Dev-iL, Jan 08 '19 at 07:42

Bar plot with two y axes

3 Answers3

Linked