MC-PCA

[CommonClimate]

Is your feature request related to a problem? Please describe.
Monte-Carlo PCA is presently disabled. We need a working one.

Describe the solution you'd like
We need:

• data structures that can accommodate multivariate ensembles (e.g. MultipleSeries made of EnsembleSeries objects)
• working, robust code, built on https://www.statsmodels.org/stable/generated/statsmodels.multivariate.pca.PCA.html
• a screeplot() function with eigenvalue quantiles derived from the age ensemble
• a modeplot() function that can map the spatial loadings, which involves resolving Geographical Coordinates in Multiple Series objects #167
• a trivial test (can repurpose tests for the pca() method)
• an interesting test, e.g. reproduce Tierney et al Nature 2013 with these records:

(this would become a PaleoHack notebook)

Could be incorporated into PCA or be a separate function.

Describe alternatives you've considered
leave things as is and let users work around the existing pca() function by looping over ensembles.

[victoriatodd2]

Is this feature available yet?

[CommonClimate]

Not yet - I was stuck on the first item, but @khider thinks it can be done with the LiPDSeries class. It's on the menu after version 1.0.0 is published, once the paper is out. If you have a more pressing need, let us know and we can collaborate!

[alexkjames]

We have some code for doing this, it will be implemented soon, waiting on #525. Once this is merged, MCPCA will be incorporated into the codebase.

[khider]

Will be tested on a Project Pythia cookbook

[khider]

Last bullet is a duplicate of LinkedEarth/PaleoBooks#11