The winter school on ‘Causal mediation analysis and machine learning’ aims at advancing the knowledge and expertise of PhD students as well as junior researchers in the field of causal inference and programme evaluation.
The first part of the two-day training session will focus on mediation analysis, which aims at disentangling a causal effect (e.g. the health effect of education) into different causal mechanisms (e.g. the health effect of education working through increased income). The second part will consider machine learning techniques aiming at leveraging big data with a large number of (control) variables for optimising outcome prediction and causal inference.
The workshop will be provided at LISER on the 3rd and 4th of February 2020, by Martin Huber, Professor and Chair of Applied Econometrics/Evaluation of Public Policies at the ‘Department of Economics’ of the University of Fribourg.
The format of the training will comprise both lectures on various methods of causal mediation analysis and machine learning as well as empirical examples using the statistical software R.
This winter school is part of the ’poverty and living conditions pillar’ of InGRID-2, a large research infrastructure aimed at integrating existing research resources in the area of ‘Advanced labour studies - Analysing the future development of work organisation, employer practices and skills analysis’ by organising mutual knowledge exchange activities and improving methods and tools for comparative research on quality of work, skills analysis and evolution of social dialogue. This integration will provide the related European scientific community with new and better opportunities to fulfil its key role in the development of evidence-based European policies for Inclusive Growth.
The winter school is organised by LISER, Luxembourg.
Location: LISER premises, 11, porte des sciences – Campus Belval, LU-4366 Esch-sur-Alzette, Luxembourg.
Start: Monday 3 February 2020, 10:30
End: Tuesday 4 February 2020, 16:30
Information for potential applicants
Requirements for attending the workshop: previous knowledge of basic principles in statistics and econometrics are required at the undergraduate level. Participation in the event is free of charge and potential participants may receive a partial reimbursement of travel and subsistence costs.
The analysis of causal mechanisms has been considered in a variety of disciplines such as sociology, epidemiology, political science, psychology and economics. This approach allows uncovering alternative causal mechanisms by studying the role of different intermediate outcomes on the causal path between the treatment and the outcome variable. The lecture on causal mediation analysis will discuss alternative strategies (eg based on observed control variables, instruments or quasi-experimental methods) to investigate such causal mechanisms. Machine learning techniques, on the other hand, are particularly suited for assessing research questions in big data contexts with many observed variables. Such methods can for instance be applied to optimally predict an outcome of interest or to optimally account for important control variables in causal inference in a data driven way.
Candidates should fill in the application form online before December 8th, 2019, including a short motivation note describing to what extent he/she would benefit from the workshop and another note about his/her current knowledge of Programme Evaluation methods (if any). Candidates will be informed at the latest by December 12th, 2019 about acceptance to the event.
Participants will be selected based on the above requirements to attend the workshop.
Participants who are scheduled to visit LISER through the visiting grants scheme sponsored by InGRID-2 will be prioritised.
Participation to the school is free of charge, and participants may be eligible for reimbursement of travel and accommodation costs. The maximum number of nights to be reimbursed equals the number of attended meeting days. Participants can find more details and rules about eligible costs as well as the reimbursement limits and procedure in the financial information of the event in question.
Participants are expected to attend the full length of the event.
Participation is subject to InGRID-2’s terms and conditions for InGRID-2 events.
Day 1: 3 February 2020
10:30 - 11:00 Welcome coffee
11:00 - 13:00 Mediation under exogeneity
13:00 - 14:00 Lunch
14:00 - 16:00 Mediation with instruments
16:00 - 16:30 Coffee break
16:30 - 18:00 Mediation based on natural experiments
18:00 - 18:30 Round table and open discussion
Day 2: 4 February 2020
08:00 - 08:30 Welcome coffee
08:30 - 10:30 Predictive machine learning
10:30 - 11:00 Coffee break
11:00 - 13:00 Causal machine learning
13:00 - 14:00 Lunch
14:00 - 16:00 Effect heterogeneity and optimal policy learning
16:00 - 16:30 Round table and open discussion
11, porte des Sciences
Esch-sur-Alzette , Luxembourg