Analysis (general)

[lilyminium]

trafficstars

Story: As a user, I want to know how whether the analysis I have in mind is already implemented, how to use it, and how to possibly implement it myself, so that I can make contributions to science.

Acceptance criteria:

• [x] I know where to find a list of accepted MD formats, and the information that each format provides.
• [x] I know where to find whether an analysis requires particular information that may not be in every MD format (e.g. masses).
• [ ] I know where to find a list of available analyses.
• [x] I know where to find a tutorial on implementing custom analysis.
• [x] I understand the general interface for running analysis.

[richardjgowers]

discussion on the Analysis interface:

https://github.com/MDAnalysis/mdanalysis/issues/719

[orbeckst]

And the current problem that the docs are missing AtomGroup methods that require masses, see https://github.com/MDAnalysis/mdanalysis/issues/1845

[orbeckst]

I would also add: the docs point to the primary literature and I know how to read about the implemented method.

Ultimately I would also need to know how to cite a certain existing analysis. The duecredit functionality might be worthwhile pointing out.

[orbeckst]

IMO, this issue seems to be the place to lead towards AnalysisBase, and not so much #7

[lilyminium]

@richardjgowers Are there any plans to refactor existing modules e.g. helanal and gnm to fit the Bauhaus style?

@orbeckst MDAnalysis gives my coarse-grained beads masses of 0 when it can't guess them from a GRO file -- is this behaviour common to all formats, or is the mass attribute simply missing from some of them?

[orbeckst]

On Sep 9, 2019, at 7:46 AM, Lily Wang [email protected] wrote:

@richardjgowers https://github.com/richardjgowers Are there any plans to refactor existing modules e.g. helanal and gnm to fit the Bauhaus style?

We would love to have it refactor if anyone would spare the time. It’s open source...

@orbeckst https://github.com/orbeckst MDAnalysis gives my coarse-grained beads masses of 0 when it can't guess them from a GRO file -- is this behaviour common to all formats, or is the mass attribute simply missing from some of them?

Good question – I don’t know. @richardjgowers? Would be good to have written down how/when we do guessing.

[richardjgowers]

I'm not sure there's a rule re: guessing masses. 0 is a good compromise for failure (though maybe np.NaN might be better?) because it's obviously not right. Sentinel values like 1 could seem correct?

[lilyminium]

Oops, sorry for the edit, I’m on my phone and thought I was replying — reverted back.

I'm not sure there's a rule re: guessing masses. 0 is a good compromise for failure (though maybe np.NaN might be better?) because it's obviously not right. Sentinel values like 1 could seem correct?

np.NaN strikes me as better because I suppose virtual sites or dummy atoms would have a mass of 0. Also, I think it might fail in mass-related calculations, which is probably desirable?