[Correl.OrdinaryLeastSquares.plot] Add option of choosing the scatter plot dot size

[BiancaMorandi]

Is your feature request related to a problem? Please describe.
When generating correlation scatter plots for different averaging periods size of dots differs depending on the number of points. It would be good if this was not the case or at least the user could have option of choosing the dot size so all plots could look the same when showing different length of data. Note that the dot auto sizing is a generic behaviour for all plot_scatter functions as it is implemented on the core _scatter_subplot function.

Describe the solution you'd like
I see two solutions:

• could add a breaking change removing the auto sizing of dots on _scatter_subplot
• could add an enhancement adding user option to _scatter_subplot of dot size and then add this option to all public functions using this subfunction and any plot_scatter functions (e.g. Correl.OrdinaryLeastSquares.plot)

Additional context
Example code showing different plots generated using Correl.OrdinaryLeastSquares.plot for different averaging periods is below with screenshot of output.

import matplotlib.pyplot as plt
import brightwind as bw
data = bw.load_csv(bw.demo_datasets.demo_data)
MERRA2 = bw.load_csv(bw.demo_datasets.demo_merra2_NE)

fig, axes = plt.subplots(2,2)

for ax in zip(axes.flatten(), ['1H', '3D', '7D','1M']):
    ord_lst_sq = bw.Correl.OrdinaryLeastSquares(MERRA2['WS50m_m/s'], data['Spd80mS'],
                                                averaging_prd=ax[1])
    ord_lst_sq.run(show_params=False)
    ord_lst_sq.plot(ax=ax[0])

[BiancaMorandi]

@stephenholleran I have added this issue as requested today.

[stephenholleran]

Thanks @BiancaMorandi.

I wouldn't classify changing the dot size as a breaking change as nothing breaks, it will just look a little different. So I would go with solution 1.

However, the reason the dots increase in size when there are so few of them is to make them more visible. If they are the same size when there are loads of data points then it is difficult to see them. The reason they are difficult to see is not just down to them been small but also because they have an opacity applied so you can tell where there is a cluster of points in the plot.

I think there are 2 real solutions here:

1. turn the dots into open transparent circles or
2. change this to a heat map type colouring

I think the open transparent circles is the way to go as it is the easiest to implement. The heat map or density plot can get tricky with sizing the area to colour in. If we do this one we would have to rely on the underlying plotting library to implement it.