PCA: output csv file for Eigenvalues and eigenvectors and facilitate plot

[Agustin Lobo]

While eigenvectors and variance ratios can be accessed via the Data Sources Panel, a more user-friendly output via optional csv files would be nice. Also, a tool to actually display the graphic of cumulated var% vs. components would facilitate the adequate selection of relevant components,

[Andreas Janz]

I guess we will not implement special functionality that only serves PCA models.

Two thinks we could implement that address your issue might be:

1. Save the whole model (like it is displayed in the Data Sources Panel) to a text file.
2. For a leave entry in the tree that represents a list of numbers (e.g. explained_variance) you can view it as a Plot (index versus value).

Would that be sufficient for now?

‌

[Agustin Lobo]

Yes, let’s try this first.

[Andreas Janz]

@Benjamin Jakimow I looked into the code and found that the original data of model parameters inside the DataSourcesTree is not available via node.value(). This just gives me a string representation with potencial “…” compression in it.

Could you make the original objects (e.g. numpy arrays) available? E.g. via node.origValue() syntax?

[Benjamin Jakimow]

As the datasource panel will be redesigned (#430) anyway, I’d prefer to not overload it with specialized optiones. Instead I’d prefer to have a widget to explore models in a much better suited way than a QTreeView (the see my comment in #520).

‌

[Andreas Janz]

on hold until the DataSourcesPanel was redesigned, which happens in v3.8

[Andreas Janz]

PCA transformer can now be browsed in the DataSources panel