Online seminar

The FOSSR Policy Learning Platform: A Tutorial

2025 gennaio 19 social card seminar GCThis webinar presents a tutorial for the Policy Learning Platform (PLP) a FOSSR tool for learning optimal policies from data allowing for empirical welfare maximization.
Implemented by the OPL set of algorithms, the PLP allows to find “treatment assignment rules” that maximize the overall welfare, defined as the sum of the policy effects estimated over all the policy beneficiaries.

The PLP allows to learn the optimal policy empirically, i.e. based on data and observations obtained from previous (same or similar) implemented policies. The PLP carries out empirical welfare maximization within three policy classes: (i) Threshold-based; (ii) Linear-combination; and (iii) Decision-tree. We present three implementation of the PLP, using three platforms, i.e. Stata, R and Python.

The seminar will be divided into three short persentations by Giovanni Cerulli (CNR-IRCrES), Federico Brogi (ISTAT) and Fabrizio De Fausti (ISTAT), followed by an open conversation with the discussants Marco Ventura (University of Rome ‘La Sapienza’) and Barbara Guardabascio (University of Perugia).


The OPL Package in Stata

The Stata OPL package is the Stata implementation of the PLP. It allows for three policy classes (Threshold-based; Linear-combination; and Decision-tree) and two different estimators of the conditional average treatment effect, i.e. the Regression Adjustment (or T-Learner), and Cross-Fitting Augmented Inverse Probability Weighting (CF-AIPW).

Giovanni Cerulli is researcher at IRCrES-CNR, National Research Council of Italy, Institute for Research on Sustainable Economic Growth. His main research interests are applied microeconometrics, with a focus on counterfactual treatment-effects models for program evaluation. He is the scientific responsible for the FOSSR project.

The OPL Package in R

OPL is an R package designed to help users learn optimal policies from data in order to maximize empirical welfare. In this session, we will showcase the key features of this innovative tool and explain how practitioners can benefit from incorporating it into their work.

Federico Brogi, 15-year experience researcher in statistics at the Methodological Division of the Italian National Institute of Statistics in Rome since 2010. His research interest includes: big data topics and policy evaluation on which he has published several articles in international journals. Since 2022, he is also a teaching assistant in Statistics at the University of Rome “Tor Vergata” and LUISS Guido Carli.

The OPL Package in Python

The presentation will illustrate the OPL software behind the PLP developed in Python, highlighting Python’s capabilities in using machine learning algorithms.

Fabrizio De Fausti, physicist, researcher since 2012 at the Methodological Division of the National Institute of Statistics. He is an expert in the development and implementation of algorithms for processing unstructured structured data. Lecturer in machine learning techniques at the Italian School of Public Administration.
 

2025 gennaio 19 locandina seminar GC
To join the event, please register at the link:
https://l.cnr.it/fossr-onlineseminar-19-feb-2025-registration-form
.
Contact person: claudia.pennacchiotti@irpps.cnr.it.

The seminar will be held online; the link will be provided to participants in due time before the beginning of the event. The seminar will be held in English, and all thematic materials will be available on Zenodo after the event.
The event is organised by CNR-IRPPS in the frame of the FOSSR project (Fostering Open Science in Social Science Research), funded by the European Union – NextGenerationEU under NPRR Grant agreement n. MUR IR0000008.

Scientific Committee: Giovanni Cerulli (CNR-IRCrES).
Organising Committee: Claudia Pennacchiotti (CNR-IRPPS), Alessandra Maria Stilo (CNR-IRCrES), Alessia Fava (CNR-IRCrES), Serena Fabrizio (CNR-IRCrES), Rita Giuffredi (CNR-IRCrES), Elisa Storace (CNR-UVR).

Data di inizio
19 February 2025 15:00

Data di fine
19 February 2025 17:00

Audience
Researchers, data scientists, PhD students

Agenda completa
Scarica PDF

Lingua
English