If your main problem is not many controls but many proxies of public transit, then I would bunch them together into 1 variable using PCA or factor analysis, as Peter suggests.

If lasso, PDS-lasso is what you want to do. You need to consider both the outcome and treatment equation. Even better: use DDML; then, you can relax sparsity and consider other learners. Stata packages: pdslasso, ddml. R: ddml, DoubleML, hdm.
statalasso.github.io thomaswiemann.com/ddml/

See how immigration numbers compare across countries: https://ourworldindata.org/explorers/migration

could you add me?

Many,many thanks to: @danielauer.bsky.social, @andreasbeerli.bsky.social, Alessandra Stambi-Bombelli, Marine Casalis, Dominik Hangartner, Chris Hansen, Selina Kurer, Stefanie Kurt, Rodrigo Sanchez, @markeschaffer.bsky.social, Michael Siegenthaler, Michaela Slotwinski, Alois Stutzer, Thomas Wiemann.

5 - Focusing again on Switzerland, we find that higher social assistance reduces criminal charges, especially for petty crimes and drug offenses.

WP: papers.ssrn.com/sol3/papers....

4 - What is the effect of cash assistance on the socio-economic integration of vulnerable Venezuelans living in Peru? We evaluate a cash-based intervention implemented by IOM and partners.

Published World Development: doi.org/10.1016/j.wo...

3 - Double Machine Learning (DML) has become hugely popular for estimating structural parameters using ML. We highlight the risk of employing ill-suited machine learners and show the benefits of pairing DML with model averaging. With R & Stata software.

Accepted @ JAE: arxiv.org/pdf/2401.01645

2 - Linking the labor market restrictions faced by refugees in Switzerland to the literature on outside options, we find that restricting refugees' outside options reduces job-to-job mobility and refugees' wages, widening the wage gap to natives in similar jobs.

WP: osf.io/preprints/so...

1 - How can we encourage eligible immigrants to naturalize? We design and field a policy tree for a letter-based information campaign. The policy tree optimally selects between alternative encouragement letters.

Forthcoming JAE: doi.org/10.1002/jae....

Could you add me? Thanks for doing this

pystacked is joint work with @markeschaffer.bsky.social
and Chris Hansen.

For more info, see statalasso.github.io/docs/pystack....