GSOC 2018 update

[josef...@gmail.com]

Google has selected the PSF as one of the mentoring organizations. Statsmodels is admitted as sub-org, for now, we still need mentor sign-up to be fully in.

http://python-gsoc.org/

Our unfinished Ideas Page for 2018 is at  https://github.com/statsmodels/statsmodels/wiki/Google-Summer-of-Code-2018

We have some higher priority projects but we are always open to "bring your own" if background of student and general need for the topic indicate a high probability of a successful GSOC project. 

(For example SVAR and sparse VAR are not listed but would make great topics for somebody interested. Dynamic panel data models are a perennial topic on the wishlist until we find a GSOC student or contributor for it.)

Josef

[josef...@gmail.com]

Another update

we have at least 3 mentors signed up with Google, so we are **in**.

I did some preliminary work and readings on parts of two of my topics categorized as "easy to intermediate" difficulty. I leave preparation for other higher priority projects to when we see a student with interest in those.

parametric survival and failure time models https://github.com/statsmodels/statsmodels/issues/4217

Post-estimation Inference and Diagnostics for GLM and discrete models

    specific to influence outliers https://github.com/statsmodels/statsmodels/issues/4268

    another part are regression plots for GLM but without detailed issue yet, those should follow regression diagnostic plots with some adjustments.

    nothing new on analysis of deviance or of pearson-chisquare some earlier experiments I did are after In69 in https://github.com/josef-pkt/misc/blob/loglinear/notebooks/ex_contingency_table.ipynb in the context of log-linear models and contingency tables

Another topic in the treatment effect estimation area that  recently showed up on the mailing list. It would be interesting to get those, most likely starting with Cattaneo and co-authors, but those might not be easy things to implement without specific background.

    

Josef