Extending Interface to Incorporate Models in Statsmodels

[jluchman]

Hi @dominance-analysis!

I am a fellow dominance analysis software developer who has mostly focused, to date, on the Stata and R software. I've been interested in your project for some time and wanted to approach you all to see if you were interested in collaborating to extend dominance-analysis' interface to generalize it in an effort to be able to incorporate most models in the statsmodels package.

Broadly, what I would propose is incorporating something like the domir R package's R formula based approach that uses a formula interface to select subsets of features but using the patsy formula structure for which many models in Statsmodels have an API.

I anticipate it would be a reasonably straightforward extension of the domir function to patsy formulas for selecting features as an interface to statsmodels' formula API. From there, it would be a matter of incorporating the architecture you all have developed in this package to implement the dominance statistic computations and dominance designations.

I would be happy to expand more on what I am thinking if this suggestion for an extension of the capabilities of the package is unclear.

Look forward to hearing from you all!

[quintshekhar]

Hi Joseph - We are willing to collaborate on this. Can we meet on a call to discuss this? Thanks, Shashank

…

On Sun, 17 Sep 2023 at 5:33 AM, Joseph Luchman ***@***.***> wrote: Hi @dominance-analysis <https://github.com/dominance-analysis>! I am a fellow dominance analysis software developer who has mostly focused, to date, on the Stata <https://github.com/jluchman/domin> and R <https://github.com/jluchman/domir> software. I've been interested in your project for some time and wanted to approach you all to see if you were interested in collaborating to extend dominance-analysis' interface to generalize it in an effort to be able to incorporate most models in the statsmodels package. Broadly, what I would propose is incorporating something like the *domir* R package's R formula based approach that uses a formula interface to select subsets of features but using the *patsy* formula structure for which many models in Statsmodels have an API. I anticipate it would be a reasonably straightforward extension of the *domir* function to *patsy* formulas for selecting features as an interface to statsmodels' formula API. From there, it would be a matter of incorporating the architecture you all have developed in this package to implement the dominance statistic computations and dominance designations. I would be happy to expand more on what I am thinking if this suggestion for an extension of the capabilities of the package is unclear. Look forward to hearing from you all! — Reply to this email directly, view it on GitHub <#35>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFLHYLV2W7NIANM6K6T2LTDX2Y45XANCNFSM6AAAAAA43GRPTU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[jluchman]

Hi Shashank,

Pleased to hear it and would be glad to join a call to talk all this over.

• joe