Estimation/validation figure

markovmodel / pyemma_tutorials

How to analyze molecular dynamics data with PyEMMA

Creative Commons Attribution 4.0 International

71 stars 34 forks source link

Estimation/validation figure #142

Closed cwehmeyer closed 6 years ago

cwehmeyer commented 6 years ago

[x] Check whether zooming improves difference visibility in (b).
[x] Missing description for line styles and grey area in (e).
[x] Text description of (f) should focus on content not interpretation.
[x] Overlap in (f) prevents black lines from being seen (this would require rewriting pyemma's plot_cktest function).
[x] Extract (e,f) as individual figure to mitigate current column height overflow.

cwehmeyer commented 6 years ago

Overlap in (f) prevents black lines from being seen (this would require rewriting pyemma's plot_cktest function).

This is a minor problem which is not easily solvable without extracting Fig.2f as a new (possibly large) figure, a new pyemma release or reinventing pyemma's plot_cktest() in NB 00.

I'd vote for skipping this part for now and providing a refactored version of pyemma.plots.plot_cktest() with the next release.

Any comment on this?

thempel commented 6 years ago

Well, I'd say that they are sort of visible if you zoom in, but I agree that's not optimal. I also can't think of a great solution here... maybe adjusting the line style of the blue line to ':' or just 'o' would be a quick and dirty fix, if that works with the API? Still not optimal though.

cwehmeyer commented 6 years ago

Well, I'd say that they are sort of visible if you zoom in, but I agree that's not optimal. I also can't think of a great solution here... maybe adjusting the line style of the blue line to ':' or just 'o' would be a quick and dirty fix, if that works with the API? Still not optimal though.

I see a minimal improve with scaling down the line widths but the API does not allow for much more control.

I would stick with line width scaling for now and address this problem later with a proper refactoring of pyemma.plots.markovtests.py (markovmodel/pyemma#1353).