Suggestion for formalizing F beta score

[shawntanzk]

Hi stato peeps,

We are trying to record confidence F scores for markers of cells in our ontology (PCL/BDSO) and was wondering if you had any suggestions on standard ways to annotate that statistic. Below is what we are trying represent:

Thanks!

[proccaserra]

Hi @shawntanzk

Thanks for reaching out. A couple of things from your use case.

• [ ] The first thing would be to have a type F-beta score to STATO.
• [ ] The second thing would be to have a way to associated a value for the beta parameter set when calculating the F-score.
• [ ] a third point would be if you'd need to know which software package was used to compute the scores, e.g. https://scikit-learn.org/stable/modules/generated/sklearn.metrics.fbeta_score.html

A question for you: could you give the reference of the book or paper your snippet comes from?

All the best

[shawntanzk]

Hi @proccaserra thanks so much for your help!

I'll be honest, I'm not great with the mathematical side of stuff, tagging @yunzhang813 here who does the NS-forest for us, she can probably help with answering some of the questions :) thanks!

A question for you: could you give the reference of the book or paper your snippet comes from?

here's the paper: https://pubmed.ncbi.nlm.nih.gov/34088715/

[proccaserra]

Thx for the pointers @shawntanzk .

Adding some material to this issue as well:

• https://github.com/JCVenterInstitute/NSForest

[yunzhang813]

Hi, here to confirm that F-beta is calculated in the NS-Forest software using from sklearn.metrics import fbeta_score to compute the score. Beta (the weight) is set to 0.5.

[proccaserra]

@shawntanzk awesome ! thx for the PR. bear with me!