Scatter plot with mean for multiple groups

Hello,

I'm trying to create a scatter plot looks like this:

I don't know the original name but I found this link : https://www.mathworks.com/matlabcentral/fileexchange/62871-cigarscatterplot?s_tid=srchtitle_scatter%20plot_15 I also attached the code that I used. It's slightly different than the original one (because I have more groups).

This is the part of data. I have 4 ranges and 3 experiments for each drug group. for every range I'm trying to add all 3 experiment for each group in same vertical area and box. eventually, my aim is to create 4 boxes for each range next to each other . (I mean; box = the area whic is seperated with dashlines in the scatterplot) and each box is needed to have 4 groups.

I wrote this command

Control= [0,0,0]
ap= [5, 36, 52]
Drug1 = [1,1,1]
bp = [32, 16, 0]
Drug2 = [2,2,2]
cp = [0, 15, 5]
Drug3 = [3,3,3]
dp= [16, 0.00, 1]
x = [Control, Drug1, Drug2, Drug3]
y= [ap, bp, cp, dp]
[x,y]= Tugce3CigarScatterPlot(x,y)

I could only add 1 range for 4 groups with all 3 experiment and looked like this. 0 is Control, 1 is Drug 1,.... , and ylim is experiment data.

I need to change 0 with Control, 1 with Drug 1,... n the xlim. Also need to add different ranges next to this plot.

I'm not sure if it's multiple questions in one question but I couldn't find any information to create this kind of plot other than the link above. If anyone knows this scatterplot's original name, maybe I can search for it and find more information, otherwise I would be really apreaciated for any suggestions for any steps.

Thank you so much

[dpb]

I can't follow the explanation of what you think the appearance should be in the end, sorry...draw a sketch by hand of what you think the result should look like for a given group/experiment to illustrate. As is, the what-you-want and what-you-got-so-far are so intertwined it's just not easy to separate...

Then, if you expect somebody to try to do something specifically towards that end, attach the data for something to work with...

[Tugce Irem]

Thank you for the suggestions. I editted my question. I added example data and "What I want" image which shows the 3 different chart should be all together and seperated with dashed lines, all need to be same sections in terms of ylim 0 to 100, xlim should be control drug1 drug2 drug3 instead 0,1,2,3 and I want to add tittle for each dashed seperated sections.

So far I get only individual sections seperately and each of them has different ylim ( 0 to 100 or 0 to 60), and I couldn't changed xlim names.

I will appreciate with any suggestions.

Thank you

Answer 1

[dpb]

fn=websave('MATLAB_exampledata.xlsx','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1136975/MATLAB_exampledata.xlsx');
tData=readtable(fn);
iscat=vartype('cellstr');
tData=convertvars(tData,@iscellstr,'categorical');
tData.Mean=mean(tData{:,vartype('double')},2)

tData = 12×6 table
     Group     Range    percentage    percentage_1    percentage_2     Mean 
    _______    _____    __________    ____________    ____________    ______

    Control    <30        98.276            100          93.103       97.126
    Control    30-39      1.7241              0          6.8966       2.8736
    Control    40<             0              0               0            0
    Drug 1     <30        89.394         87.692              90       89.029
    Drug 1     30-39      10.606         12.308              10       10.971
    Drug 1     40<             0              0               0            0
    Drug 2     <30        61.765          75.61             100       79.125
    Drug 2     30-39      36.765          24.39               0       20.385
    Drug 2     40<        1.4706              0               0       0.4902
    Drug 3     <30        33.962         1.6129          88.043       41.206
    Drug 3     30-39       52.83         43.548          11.957       36.112
    Drug 3     40<        13.208         54.839               0       22.682

hTL=tiledlayout(1,3);
R=categories(tData.Range);
for i=1:numel(R)
  hT(i)=nexttile;
  if i==1, posn=hT(i).Position; L=posn(1);B=posn(2);H=posn(4);end   % left, bottom, height
  ix=(tData.Range==R(i));

  hSC=scatter(tData(ix,:), 'Group', {'percentage', 'percentage_1', 'percentage_2'},'filled');
  hold on
  hL=plot(double(tData.Group(ix)).'+0.5*[-1;1],double(tData.Mean(ix)).'+[-0;0], 'r-');
  ylim([0 100]+[-2 2])              % add just a little whitespace 
  hT(i).XAxisLocation='origin';     % and keep axis at origin
  title(R(i))
  if i>1, hT(i).YAxis.Visible='off';end
  xlabel('')
  if i==numel(R)
    hLg=legend([hSC(:);hL(1)],'Pct 0','Pct 1','Pct 2','Mean', ...
                           'Location','eastoutside','Orientation','vertical');
    posn=hT(i).Position; W=posn(1)+posn(3)-L;  % width overall after adding legend
  end
end  
hA=annotation('rectangle',[L B W H]);

Would be my starting point with a traditional scatter can't run the cigarScatter on online platform; I'll see if I can find some time to look at it some later...that's most of what you wanted on the axes; it isn't so easy to draw the horizontal line with categorical variables; yline works but goes across the whole axis range so that doesn't help and there's only the one point on a categorical axes at which can plot. One way is to introduce "fake" categories into the variable to introduce more granularity in the axis. Another is to lay another axis on top the the existing that is numeric and plot into it; alternatively, just do as the cigar plot guy did; do it all numerically and the just label the ticks; the tick labels and the ticks are independent properties.

Mayhaps that will give some ideas on approaches...

ADDENDUM: "Stealing" from @Chunru using the table variable names and the new feature (introduced R2022b) of being able to use other than a categorical variable in a categoricalRuler axis, just a few enhancements.

Also to add the overall figure box is not too bad with the annotation function; just have to compute the outer boundaries of the axes from first to last to find the overall coordintes.

To add the dividing lines between is another application of annotation with a line; one needs to add the computation of the x-location between the first two and the last two axes to come up with the position at which to draw it; I've left that as "exercise for student" -- it would follow the idea illustrated above in the computation of the overall width -- one has to find the RH position of the first axis by adding the Left and Width position values and then add halve the distance between that location and the Left postion of the second.

With the advent of being able to use numeric data on categorical axis, all it would take would be to add a little jitter to the x position vector to be able to create the "cigarscater" figure without having to fool with it. I looked at it a little last night but it had so much extra baggage coming along with it to handle all the "what ifs" that it does that I didn't feel like trying to delve into just how it actually worked to try to do anything more with it.

[Tugce Irem]

Thank you so much for your detailed explanation. I'll try your suggestions.

[Chunru]

@dpb Nicely done.

[dpb]

@Chunru -- thanks.

[Tugce Irem]

Yeeeyyyyy thank you so so muchhh :) @dpb @Chunru

[Tugce Irem]

@dpb I tried but couldn't do that , sorry need to look at how I can accept multiple answers because there is no accept button left after I choosed one. But I really appreciate to both of you (@dpb and @Chunru)

May I ask one more thing, I want to move down xlabel ? Or should it be a new question? When I write something longer than drug1, my words came in to the box. I added some space or space and dot as a name in the excell ( Drug1 , or Drug1 .) these are looks weird.

also changed this;

hT(i).XAxisLocation='origin' , 'Location','southoutside';     % and keep axis at origin  

I can write a new question if necessary, and thank you again I've already asked to many question.

[dpb]

Hmmm...mayhaps I was/am wrong; I thought one could do so with the more recent changes -- I know it doesn't cost a poster status points if change, though.

On the xlabel position, if you want to move it, then save its handle when creating it and adjust the 'Position' value...it's just a text object with (x,y,z) coordinates (z-->0 for 2D graphs)

hXLab=xlabel('X');          % write a label; save handle
posn=hXLab.Position;        % retrieve default position
posn=posn.*[1 1.1 1];       % adjust y position by 10% (will be negative so down)
hXLab.Position=posn;        % set new position

On axis location,

hT(i).XAxisLocation='origin' , 'Location','southoutside';     % and keep axis at origin  

is mixing metaphors of the legend positon with the axis -- they're entirely different; the axis location is only one of 'origin', 'top', 'bottom'; anything else is unallowble.

It's some other modifications up earlier to be able to put the legend at 'southoutside'; I actually had tried it and decided didn't want to take the time to make the other modifications to do so, so took the easiest way out.

To do that and make it all work out in the end, you'll have to add it first before calculating the location of the bottom axes because adding it causes the axes to all be shrunk to be shorter to make room. That'll mean moving the logic up top inside the loop that computes L, B, H to after the legend is created in order to have the axes final positions.

It's all doable; just a lot of bookkeeping and nits to take care of to create the special effects desired. Welcome to HG2!!! You'll be a whizard here shortly and be complaining like me about stuff that should be simpler than it is...

Answer 2

[Chunru]

x = readtable("https://www.mathworks.com/matlabcentral/answers/uploaded_files/1136975/MATLAB_exampledata.xlsx");
x.Group = categorical(x.Group);
x.Range = categorical(x.Range)

x = 12×5 table
     Group     Range    percentage    percentage_1    percentage_2
    _______    _____    __________    ____________    ____________

    Control    <30        98.276            100          93.103   
    Control    30-39      1.7241              0          6.8966   
    Control    40<             0              0               0   
    Drug 1     <30        89.394         87.692              90   
    Drug 1     30-39      10.606         12.308              10   
    Drug 1     40<             0              0               0   
    Drug 2     <30        61.765          75.61             100   
    Drug 2     30-39      36.765          24.39               0   
    Drug 2     40<        1.4706              0               0   
    Drug 3     <30        33.962         1.6129          88.043   
    Drug 3     30-39       52.83         43.548          11.957   
    Drug 3     40<        13.208         54.839               0   

urange = unique(x.Range);
n = length(urange);
for i=1:n
    subplot(1,n,i)
    x1 = x(x.Range==urange(i), :);
    scatter(x1, 'Group', {'percentage', 'percentage_1', 'percentage_2'});
    hold on
    plot((double(x1.Group)+[-.5 .5])', (mean(x1{:, 3:5}, 2)+[0 0])', 'r-')
    title(urange(i))
end

[Tugce Irem]

Super thank you both @dpb and @Chunru. I'll try to combine both your answers since @dpb showed the data in all same yline and you add mean to the charts. Thank you so much again :)

[dpb]

@Chunru -- where did you find the facility in scatter to use grouping variables? I can find no reference to that in R2022b doc's -- I had forgotten about gscatter in the Statistics TB and just discovered above by accident that base scatter has been upgraded from vector x,y only to accepting a matrix (I'm still on R2020b on base machine; these came R2021 or later).

It's also a revelation about the addition of the line to what look to be categorical axes -- unless they're not in using the new version; I'll have to poke around some at the innards of the axes object...

[Steven Lord]

In this case 'Group' is not a name-value argument, it's the name of a variable in the table. If you look at the documentation page for the scatter function this is using the scatter(tbl, xvar, yvar) syntax. See the "Plot Data from a Table" example on that page. This functionality was introduced in release R2021b.

[dpb]

@Steven Lord -- (Headslap!) Yeah, thanks! I got myself confused and then couldn't get the wrong idea out of head. Too used to scatter not being vectorized to take an array, I guess.