logarithmic density estimate

[KevinAquino]

Hi all,

First off, love the repo!

Secondly, I was wondering what types of features can help with visualising the tails of distributions? I currently have three datasets with different tails, the "drops" do a nice job of visualising the tails, but I feel like the default density estimate is lacking in that regard - I would typically analyse the density estimate with a logarithmic scaling (see second plot)

Are there features in the pipeline to deal with this (now or in the future)? It makes the raincloud plot a little different, but maybe this has been thought of already?

[pog87]

Hi Kevin.

Thanks for writing us!
Did you try to change the bw input argument?

[KevinAquino]

Hiya Davide,

I'm using the Matlab implementation so I assume you mean changing the band_width of the KSD estimate? I tried changing this but the density plots seem to converge at the tails. The KSD estimate is not too bad but visually I think a logarithmic y-axis would useful to add to the standard raincloud plot - although it would be a log-rain-cloud plot.

The distinction could be made clear if there was a y-axis displayed so one knows if we have a logarithmic or a standard rain-cloud plot, below is an example of a "log-rain-cloud" plot

[pog87]

I see your point, this is a choice one user could look for, as in the case of log-histograms. Can you show some code you used to make the logCloud?

[micahgallen]

This does sound like something that could be an easy across-system add-on. Once we see the python implementation we can discuss adding more generally.

On Mon, 3 Jun 2019 at 14.23, davide poggiali ***@***.***> wrote: I see your point, this is a choice one user could look for, as in the case of log-histograms. Can you show some code you used to make the logCloud? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#31?email_source=notifications&email_token=AKDTU5BOEMHESVQ72TGRQYLPYUEMTA5CNFSM4HRROCZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODWZHIDY#issuecomment-498234383>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKDTU5DVNQN4TFJWG4E3WSTPYUEMTANCNFSM4HRROCZQ> .

-- Dr. Micah Allen Associate Professor @ AIAS <http://aias.au.dk/> & Senior Research Fellow @ Cambridge Psychiatry <http://www.psychiatry.cam.ac.uk/> micahallen.org +44 07473513511

[KevinAquino]

Sorry for the delay, was writing this just before a conference!

Here is a section of code I hacked together rather quickly to do this: This is clunky 🙈 and fit for my purpose in Matlab. It is tricky to get the logarithmic axes working so the box plot with the rain drops need to be made in a different axis to the cloud.

In action its in a latest ms I've submitted : https://www.biorxiv.org/content/10.1101/662726v1 figure 12

% Generate some synthetic data:
data{1} = randn(1000,1);
data{2} = randn(1000,1);
data{3} = randn(1000,1).^3;



figure('color','white');
ax1 =subplot(211);
% Generate some colors for the mapping
theColors{1} = [0 0 1];
theColors{2} = [0 1 1];
theColors{3} = [1 0 1];
% Label these data
labels={'one','two','three'}

% Make generic Raincloud plots
h1 = raincloud_plot(data{1}, 'box_on', 1, 'color', theColors{1}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .15, 'dot_dodge_amount', .15,...
     'box_col_match', 0);

h2 = raincloud_plot(data{2}, 'box_on', 1, 'color', theColors{2}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .35, 'dot_dodge_amount', .35, 'box_col_match', 0);

h3 = raincloud_plot(data{3}, 'box_on', 1, 'color', theColors{3}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .55, 'dot_dodge_amount', .55, 'box_col_match', 0);

% Turn off the box plots
for pl=2:6
	set(h1{pl},'visible','off');
	set(h2{pl},'visible','off');
	set(h3{pl},'visible','off');
end

set(gca,'fontSize',18,'Ylim',[1e-8 1],'XLim',[-5 10],'YScale','log','XTick',[]);
ylabel('Density');

ax2=subplot(212);

% replicate the rain cloud plots
h1rd = raincloud_plot(data{1}, 'box_on', 1, 'color', theColors{1}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .15, 'dot_dodge_amount', .15,...
     'box_col_match', 0);

h2rd = raincloud_plot(data{2}, 'box_on', 1, 'color', theColors{2}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .35, 'dot_dodge_amount', .35, 'box_col_match', 0);

h3rd = raincloud_plot(data{3}, 'box_on', 1, 'color', theColors{3}, 'alpha', 0.5,...
     'box_dodge', 1, 'box_dodge_amount', .55, 'dot_dodge_amount', .55, 'box_col_match', 0);

ylabel('Box plots');
set(gca,'YLim', [-0.35 0],'XLim',[-5 max(data{3}(:))*1.1],'fontSize',18,'YTick',[]);

% Turn off the clouds!

set(h1rd{1},'visible','off');
set(h2rd{1},'visible','off');
set(h3rd{1},'visible','off');

legend([h1{1} h2{1} h3{1}], labels);

ax1.OuterPosition = [0 0.45 1 0.5];
ax2.OuterPosition = [0 0 1 0.5];