Add dots at peak of bars in a bar graph and draw a line between them?

Hi all,

I originally plotted data in a bar graph with brain regions along the x-axis and some measure along the y-axis. Each region has 2 data points, one for each condition. I have since realized that a bar graph is not the best way to visualize said data; instead, I want to draw dots at the peak of each bar and a line connecting the two dots of a single region (one for each condition). From the attached figure, that would mean that at RH PHC2 for example, there would be a blue dot at the peak of the left/blue column and an orange dot at the peak of the right/orange column with a line connecting the two dots.

I assume that the easiest way to accomplish this would be to add the dots and lines to the current graph, and then make the bars transparent resulting in a pseudo-dot plot. I don't know how to do this however. I've attached the code of my figures thus far in hopes that I can receive some assistance.

If it is easier to start from scratch, I'm happy to do that too, but would appreciate getting pointed in the right direction!

Thanks,

George

>> BetRSF

BetRSF =

13.0000 0 277.9367 347.5332 6.0000 1.0000

48.0000 0 245.8306 315.4521 24.0000 2.0000

26.0000 0.0168 133.7187 141.7139 12.0000 3.0000

49.0000 0 109.6015 133.0241 25.0000 4.0000

46.0000 0 85.4435 101.6307 22.0000 5.0000

36.0000 0 69.4742 74.8517 17.0000 6.0000

30.0000 0 63.1806 68.0751 14.0000 7.0000

>> RegRSF

RegRSF =

7×1 categorical array

RH PHC2

RH vPCu

RH TO1/MT

RH vTO

RH SEF

RH IPS3

RH V7/IPS0

[BRSFrow, BRSFcol]=size(BetRSF); % need rows for 'for' loop

figure(15)   
%     subplot(3,2,4)
        bar(RegRSF,BetRSF(:,3:4))
        ylim([0 370]);
        title('Right - Saccade Greater');
        for i=1:BRSFrow
            text(i-.05, BetRSF(i,4)+10, '*','FontSize', 20);
        end

Answer 1

[Adam Danz]

Working with categorical axis rulers is tricky because you can't use simple arithmetic to get the bar centers. The demo below compares the bar plot to a line plot. You can't use categorical x values if you want two markers with slight separation per category. This solution uses a numeric x axis ruler and uses the category names as XTickLabels. If you want he categories to be in the same order as the bar plot, you have to rearrange the data to match the categorical order which this solution also demonstrates. Lastly, the separation between groups is set to 0.1 which is arbitrary but affects the slope of each line segment.

BetRSF =[
    13.0000         0  277.9367  347.5332    6.0000    1.0000
    48.0000         0  245.8306  315.4521   24.0000    2.0000
    26.0000    0.0168  133.7187  141.7139   12.0000    3.0000
    49.0000         0  109.6015  133.0241   25.0000    4.0000
    46.0000         0   85.4435  101.6307   22.0000    5.0000
    36.0000         0   69.4742   74.8517   17.0000    6.0000
    30.0000         0   63.1806   68.0751   14.0000    7.0000];

RegRSF = categorical([
    "RH PHC2"
    "RH vPCu"
    "RH TO1/MT"
    "RH vTO"
    "RH SEF"
    "RH IPS3"
    "RH V7/IPS0"]);

figure()
ax(1) = subplot(2,1,1);
h = bar(ax(1), RegRSF,BetRSF(:,3:4));
grid(ax(1),'on')

ax(2) = subplot(2,1,2);
grid(ax(2),'on')
hold(ax(2),'on')
box(ax(2),'on')
x = (1:numel(RegRSF)) + 0.1*[-1;1]; % set inter-group distance
y = BetRSF(:,3:4)';
% Reorder y to same order as barplot
[~, idx] = ismember(categories(RegRSF),cellstr(RegRSF));
y = y(:,idx);
% Plot lines and endpoints
lh = plot(ax(2), x,y, 'k-','LineWidth',2);
plot(x(1,:), y(1,:), 'ko','MarkerFaceColor','b')
plot(x(2,:), y(2,:), 'ko','MarkerFaceColor','r')
xlim(ax(2), [.5, numel(RegRSF)+.5])
ax(2).XTick = 1:numel(RegRSF);
ax(2).XTickLabel = categories(RegRSF);
ax(2).XTickLabelRotation = 25;

[George Tomou]

The categorical x-axis has been giving me the hardest time in terms of plotting the dots and lines. I was working on this and was coming up with either having the categorical axis but having both conditions plotting on the same vertical, or having to designate my own x-axis. I couldn't find how to get both options in one plot. This works perfectly - thanks!

[Adam Danz]

Glad it work out for ya!

What's wrong with your first subplot where the blue and red values are stacked on top of eachother? That gives the best visual comparison of the two values. Although I suppose the connector line helps with visual comparison, too.

[George Tomou]

I agree - there's nothing wrong with the first one per-se. I'll be presenting both and seeing which my colleagues agree with. Thanks again!

[Adam Danz]

Make sure you include a legend 😎

[George Tomou]

Hi again Adam,

Hope you get a notification for this after so long.

This is pretty close to the final version of this figure, but my supervisor wants to "trim the excess white space". Is there a way to do this by setting the space between the ticks on the x-axis such that the figure is reduced in size overall?

[Adam Danz]

axis tight
% or
axis(h,'tight') % h is axis handle

or, if you're using Matlab r2020b or later,

axis padded
% or
axis(h,'padded') % h is axis handle

Or just set the ylim

ylim(h, [50,360])

[George Tomou]

Hmm I may have miscommunicated my question in the previous comment. I have edited it now so that hopefully it clarifies the confusion.

In the code, I have set the ylim. What I'd like to do now, is reduce the spacing left to right. Essentially, "tightening up" the x-axis spacing.

I try using axis(h, 'padded') but it does not seem to work. As is, I get the following error which might be because I'm not using the correct axis handle?

Error using axis>LocSetLimits (line 323)
Vector must have 4, 6, or 8 elements.

Error in axis (line 114)
                LocSetLimits(ax(j),cur_arg,names);

Here's the code:

BetRSF =[
    13.0000         0  277.9367  347.5332    6.0000    1.0000
    48.0000         0  245.8306  315.4521   24.0000    2.0000
    26.0000    0.0168  133.7187  141.7139   12.0000    3.0000
    49.0000         0  109.6015  133.0241   25.0000    4.0000
    46.0000         0   85.4435  101.6307   22.0000    5.0000
    36.0000         0   69.4742   74.8517   17.0000    6.0000
    30.0000         0   63.1806   68.0751   14.0000    7.0000];

RegRSF = categorical([
    "RH PHC2"
    "RH vPCu"
    "RH TO1/MT"
    "RH vTO"
    "RH SEF"
    "RH IPS3"
    "RH V7/IPS0"]);

ax(2) = subplot(1,2,2);
grid(ax(2),'on')
hold(ax(2),'on')
box(ax(2),'on')
x2 = (1:numel(RegRSF)); % no inter-group distance
y2 = BetRSF(:,3:4)';
% Reorder y to same order as barplot
[~, idx] = ismember(categories(RegRSF),cellstr(RegRSF));
y2 = y2(:,idx);
% Plot error bars %
yrfix=y2(1,:);
yrsac=y2(2,:);
err1=errorbar(x2,yrfix,errRF,'k.','CapSize',5,'LineWidth',1.1);
err2=errorbar(x2,yrsac,errRS,'k.','CapSize',5,'LineWidth',1.1);
% Plot lines and endpoints
% lh = plot(ax(2), x,y, 'k-','LineWidth',2);
p1=plot(x2(1,:), y2(1,:), 'bo','MarkerFaceColor','b');
p2=plot(x2(1,:), y2(2,:), 'ro','MarkerFaceColor','r');
title('Right Hemisphere');
legend([p1 p2],'Fixation','Saccade')
xlim(ax(2), [.5, numel(RegRSF)+.5])
ylim(ax(2), [0 370])
axis(x2,'padded') 
    % not sure if x2 is the correct variable for axis
    % handle
ax(2).XTick = 1:numel(RegRSF);
ax(2).XTickLabel = categories(RegRSF);
ax(2).XTickLabelRotation = 25;

[Adam Danz]

There isn't much space on the right and left side of the axes but to reduce it, you can use the same advice in my previous comment.

Since the x-axis is categorical, xlim() won't work unless you want the axis to start and stop exactly on the first and final categorical tick.

I'd go with axis tight or axis padded (padded requires Matalb r2020b or later).

More importantly, if you're providing a handle, which is advised, it should be an axis handle which does not seem to be the case in your comment above.

axis(x2,'padded') 

x2 appears to be a numeric vector.

Also your Release info statese that you're using r2020a. Perhaps you didn't read my comment close enough.

If axis tight isn't good enough your options are limited in r20a.

[George Tomou]

I switched to 2020b last night following your advice on my heatmap thread and never updated this one. Updated it now to reflect the change.

[Adam Danz]

Well, in that case the padded option should work all you need to do is make sure the handle is the axis handle. On second thought, I haven't tested it with categorical axes so let me know how it works out.

[George Tomou]

It looks like it works with categorical data. I grabbed the right handle - axis(ax(2).XLimMode, 'padded'). Good news, it "tightens up" the first and last data points by bringing in the figure border in tighter, but does little to the spaces between the remaining data points. Guess that's just a matter of manually adjusting the figure window after figure generation.

Thanks again, as always!

[Adam Danz]

It should be

axis(ax(2), 'padded')

To narrow the spaces between points, you can change the axis size

ax(2).Position(3) = ax(2).Position(3).*.75; % 75% the width

or the figure

fig.Position(3) = fig.Position(3)*.75;  % 75% the fig width

Answer 2

[Sumedh Deshpande]

Hi George,

I am assuming are using R2020a or higher version of MATLAB.

Here is one way of getting the graph with lines, just comment the line hold on and hold off to get just the line graph

[BRSFrow, BRSFcol]=size(BetRSF); % need rows for 'for' loop
figure(15)
%     subplot(3,2,4)
bar(RegRSF,BetRSF(:,3:4))
ylim([0 370]);
title('Right - Saccade Greater');
forLine = [];
for i=1:BRSFrow
    text(i-.05, BetRSF(i,4)+10, '*','FontSize', 20);
    forLine = [forLine  BetRSF(i,4)+10]; % to store the locations of * in a an array
end

hold on; % to keep the old graph 
plot(forLine) % plot the lone graph
hold off;