PVD Plotting Update

[sea-shunned]

After a few requests for tweaking parts of plots, made a few changes to update the PVD plotting code. Changes summarized below:

• For all PVD plotting functions (*._plot_sustain_model):
  • Added optional arguments control font size (and rotation) for all elements (title_font_size, stage_font_size, label_font_size, stage_rot, label_rot)
  • Added option to control xtick interval (stage_interval) for when there are many events
  • Optional argument for figure size (figsize)
  • Optional argument to colourize biomarker labels (biomarker_colours). Colours must pass matplotlib.colors.is_color_like, i.e. be a name or RGB(A).
    • A dict maps label:colour. Any labels not given are defaulted to black.
    • A list/tuple presumes default order of labels, and must all be given.
• plot_format option added to run_sustain_algorithm and combine_cross_validated_sequences to specify saving format
  • Both also pass all kwargs to _plot_sustain_model
• MixtureSustain
  • Fixed colourmap, so can now be added. Limits defined to handle edge case of all <1 values passed, allowing visual comparison with other PVDs
• ZscoreSustain
  • Added missing implementation for biomarker_order
  • Upto 5 z-scores can be given. Use of arbitrary colourmap potentially problematic (as noted in a comment), and not visually informative. If more z-scores are desired, custom solution can be added.
• ZscoreMissingSustain
  • Just returns ZscoreSustain version

[Zeena-Shawa]

Works perfectly when trying to plot 6 subtypes, but won't work for plotting 4 or 5 subtypes. When trying to plot 5 subtypes, the first three are correctly plotted, but then it exits with an error: IndexError: index 5 is out of bounds for axis 0 with size 5. For reference: plot_subtype_order = [1,2,0,5,4,3]; samples_sequence = 5; N_S_max = 6;

[sea-shunned]

Thanks @Zeena-Shawa

Seems that this occurs when you are create a model (e.g. MixtureSustain instance), but then try to plot a PVD for a different number of subtypes than N_S_max with variables (samples_sequence etc.) loaded in from pickle files.

Possible solutions are:

1. Override the internal self._plot_subtype_order if there is a shape mismatch with samples_sequence. You can either use an arbitrary subtype order in that case, or reconstruct using the original approach (np.argsort(ml_f_EM)[::-1]), thus needing to add that as an extra argument.
2. Turn the plotting function into a @staticmethod, so that self isn't needed at all. This would break backwards compatability, but massively increase the flexibility of the plotting function, by no longer requiring an instance (and thus all that is needed is to load pickle files, rather than recreate the instance, which is useful when the compute has been off-loaded to HPC). I prefer this, but it will require people to slightly edit old code.
3. Just raise an error when there is a shape mismatch, and advise that the subtype_order argument is used manually. Least preferred option, as this is not the path of least resistance for the user.

Thoughts welcome!

[noxtoby]

I vote for number 2 — I have no problem with breaking backward compatibility, given that we have no specific releases (and aren't in a release cycle, for example).

[sea-shunned]

Great, that's 2 people, and by the power of democracy that'll do.

[sea-shunned]

Went with option no. 4: public-facing generic function (*.plot_positional_var) that can be used with pickle_files, that is also used by the original self._plot_sustain_model to maintain backward compatibility...so best of both worlds (in theory!).

Tested Mixture & Zscore versions with the simrun.py, but not Ordinal and ZScoreMissing yet.

[Seymour22]

The title_font_size, stage_font_size parameters are working when passed in ._plot_sustain_model, but figsize doesnt work. None of these are recognized in combine_cross_validated_sequences though. Maybe you can help clarify a install question. Where do you run the git switch plotting command. I've done in in the sustain env, and within your downloaded folder (pySuStaIn-plotting). As I kept getting git switch is not a command* error, I just ran the pip install . from pySuStaIn-plotting

[Seymour22]

Made these changes in the scripts...still doesn't recognise figsize parameter

[sea-shunned]

Thanks for testing @Seymour22. Replies below:

• figsize: It works for me (both zscore and mixture versions), can you provide more detail?
• combine_cross_validated_sequences: Passing arguments here also works for me.
• git switch: It's an alternative to git checkout that requires git v2.23+. git checkout plotting will do the same thing (see here for more info).

[Bea-Taylor]

Worked great with ordinal sustain, tried out the different arguments and all working well 💯 Thanks Cameron!

[sea-shunned]

Great to hear @Bea-Taylor, and thank you for testing it out!

Edit: I'll add a docstring to plot_positional_var to explain its usage, though long-term we need to modify the existing notebooks and simulation scripts to use these new functions, demonstrating how they are used (the two main scenarios interfacing with plot_positional_var and run_sustain_algorithm.

[williamscotton]

Hey Cameron this is really useful addition! Was wondering whether it would be useful to have ability to plot each subtype PVD as a separate figure e.g. 3 subtypes currently plotted all together on one plot - have a argument in the plotting function that allows you to output each plot as a separate figure i.e. 3 separate figures for a 3 subtype model

[williamscotton]

..Also do you think you might be able to add a plotting funciton that allows you to plot number but stage - potentially have an argument that allows you to select subtype 1 by stage etc then an option that plots all subtypes by stage on one plot with each subtype colour coded?

[williamscotton]

Hey Cameron this is really useful addition! Was wondering whether it would be useful to have ability to plot each subtype PVD as a separate figure e.g. 3 subtypes currently plotted all together on one plot - have a argument in the plotting function that allows you to output each plot as a separate figure i.e. 3 separate figures for a 3 subtype model

and also to be able to save figure for each at time of running to folder!

[sea-shunned]

Hey Cameron this is really useful addition! Was wondering whether it would be useful to have ability to plot each subtype PVD as a separate figure e.g. 3 subtypes currently plotted all together on one plot - have a argument in the plotting function that allows you to output each plot as a separate figure i.e. 3 separate figures for a 3 subtype model

Sure, sounds good! I'll add a separate_subtypes flag for this, in which case a list of figures and axes objects will be returned.

and also to be able to save figure for each at time of running to folder!

Great point, plot_positional_var as a standalone function should have some save functionality, which will save the whole figure or individual ones depending on the new option above. I'll add options for save location, save name, and extra kwargs for fig.savefig.

Thanks @williamscotton for taking a look and your suggestions!