Add quantile feature to forestplot ridgeplot

[percygautam]

Description

Fixes #756 . Add new feature quantiles to ridgeplot of forestplot.

Checklist

• Does the PR follow official
  PR format?
• Has included a sample plot to visually illustrate the changes? (only for plot-related functions)
• Is the new feature properly documented with an example?
• Is the code style correct (follows pylint and black guidelines)?
• Is the change listed in changelog?

[percygautam]

Currently implemented for matplotlib. It looks something like:

There is a problem of having step size too small in plots. ax.fill_between will inverse fill between x[i] and x[i+1], this cause the quantile line to be very thin. (almost disappear in more dense step sized plots like defs Italy). How can we clearly show these quantiles?

[aloctavodia]

As a workaround you can stretch the figure along the x-axis, something like figsize=(14, 4). Another solution is to re-think how we display quantiles for kdes. Maybe instead of not filling the area we could use two alternating shades (using two different alpha values) ¿?

[percygautam]

@aloctavodia I tinkered a little with figsize generation, and below size seems okayish:

    if figsize is None:
        if ridgeplot_quantiles is None:
            figsize = (min(12, sum(width_ratios) * 2), plot_handler.fig_height())
        else:
            figsize = (min(12, sum(width_ratios) * 5), plot_handler.fig_height() * 1.2)

Graph looks like this :

About alternate shades for quantiles, Should I add that also in this? Should alternate shades also be added to normal kdeplot with quantiles?

[aloctavodia]

Not sure, but I guess a good solution is to work on this first and after merge work on a separated PR for the shade feature. For the next PR we may even think on a general function that is called by the kdeplot and the ridgeplot.

[aloctavodia]

I do not think the function should change the figsize automatically, the user should be in control of the figsize

[percygautam]

Okay, Let's get this one done first. I have changed the figsize for ridgeplot_quantiles, if it is not explicitly defined by the user. It's creating a figure of (12, 5.2). Is this correct or should we specify some different default figsize for quantiles plot?

[percygautam]

I have added the quantile feature to bokeh ridgeplot, and changed the default figsize to clearly show the quantiles (for [0.2, 0.8]):

Also, The figsize of normal ridgeplot (without quantiles) is not normalized, making the bokeh figure to look cluttered. Should I change that also (just for ridgeplot)?

[aloctavodia]

ohh, sorry I misread your previous example. This is OK

[ahartikainen]

LGTM

[ahartikainen]

this could be

xvals = yvals = pdfs = pdfs_q = colors = []

but it is not that important, so no need to change it