Professor of Applied Econometrics and Policy Evaluation at
@ses_unifr
@unifr
- causal analysis, statistics, econometrics, machine learning...and telemarking
News concerning my book
#CausalAnalysis
(): The R and Python code along with the datasets in CSV format for the empirical examples featured in the book are now also available on the Harvard Dataverse repository:
#Rstats
#Econtwitter
Our working paper on
#MachineLearning
for staggered
#DifferenceInDifferences
is out! Our method explores effect heterogeneity under staggered treatment adoption and is applied to assess the effect of Brazil's Family Health Program on infant mortality:
A new version of the causalweight package for
#RStats
is online:
Now including example datasets for the regression kink design and the estimation of interference/spillover effects in treatment evaluation as considered in my book
#CausalAnalysis
#Econtwitter
If you got lost in the jungle of recent developments in
#causalanalysis
based on difference-in-differences, you should get this (free) book (work in progress) by
@CdeChaisemartin
and Xavier D'Haultfœuille: 👍😀
#CausalInfernce
is spreading across all parts of society 😀 The target group of this book 👇by Vasil Yasenov are...toddlers (and probably other non-statisticians with an interest in cause and effects). You can't start too early with learning about
#causality
😉
Take a look at our updated causalweight package for
#CausalAnalysis
in
#RStats
: . It now includes a test for the identification of causal effects in observational data as proposed in our working paper: .
#EconTwitter
#CausalTwitter
😀 Just learned that my book
#CausalAnalysis
is a finalist for the "Association of American Publishers Awards for Professional and Scholarly Excellence" (PROSE Awards) in the category "Computing and Information Sciences":
A very interesting & shocking diff-in-diff-based
#CausalAnalysis
finds that
@Nestle
's infant formula market entrance in lower-income countries substantially increased infant mortality by more than 20% (!!) among those using unclean water:
#CausalTwitter
Stumbled on a staggering result:
Starting in the 1960s,
@Nestle
aggressively marketed baby formula > breastfeeding in LMICs. Formula was often mixed with unclean water
The result: increased infant mortality and *212,000* excess deaths per annum by 1981
My survey on causal discovery (learning causality in a system of variables in a data-driven manner) is now available on ArXiv, as pointed out by
@CavaliereGiu
:
Hi
#EconTwitter
! 📊
Interested in causal inference?
Check out this brand new survey by
@CausalHuber
on "causal discovery" — where data-driven methods are used to find causal relationships between many variables. 📚
Don't miss out — very interesting stuff! ⭐️
Link:
🔥Our new working paper (joint with
@TotaroApfel
,
@HatamyarJ
, and J. Kück) proposes a data-driven method for identifying (sets of) variables that satisfy either selection-on-observables or instrumental variables assumptions in causal models:
#EconTwitter
😀Check out my new data repository on "Impact Evaluation in Firms and Organizations," featuring datasets and
#RStats
/
#Python
code for business use cases in impact evaluation (eg, assessing the sales impact of marketing interventions):
#DataScience
Delighted to present our paper (with J. Kueck) "Testing the identification of causal effects in observational data" in the online econometrics seminar at
@BrownUniversity
today. Excited to engage with faculty members such as
@jondr44
and
@instrumenthull
.
In social sciences,
#CausalAnalysis
mostly focuses on evaluating a treatment, as seen in .
#CausalDiscovery
(prominent in
#computerscience
) aims to learn causality within an entire system of variables - here's a gentle introduction:
Prof.
@yudapearl
leading the speakers session in this year's Causal AI conference along with Guido Imbens!
Are you attending in person, Prof.
@yudapearl
?
It is wrong that IV assumptions for
#causalinference
are not testable. See here for some of the tests recently proposed:
You can even combine testing with
#machinelearning
:
🔥Update of our paper Testing the identification of causal effects in observational data (joint with J Kück), providing a
#MachineLearning
test for the joint satisfaction of selection-on-observables & IV validity, which is sufficient for identification:
Our new publication (joint with Y. Hsu, Y. Lee, L. Lettry) on
#causalmediation
analysis for continuous treatments (disentangling direct and indirect causal effects). Code available in the causalweight package for R.
#econtwitter
#RStats
#causalinference
In IV regression, converting multivalued treatments (eg education) to binary ones(high/low education) may violate the
#exclusionrestriction
. Our publication (with
@eckhoffandresen
) provides conditions for identification+testable implications:
#Econtwitter
🔥Finally published in the
#Econometrics
Journal: Our paper (joint work with Christian Dahl and
@GioMellace
) on the instrument-based evaluation of local average treatment effects in the presence of defiers (non-monotonicity):
We've updated our paper on estimating local average treatment effects (LATE) using multiple instruments when only some of them are valid. Our method detects the valid instruments (under certain conditions) & allows exploring LATE heterogeneity
#EconTwitter
Special thanks go to
@causalinf
and
@InstEconomist
for their great support in the process of finding a publisher,
@mitpress
for publishing the book, and to the following colleagues who took the time to read the pdf draft version and made very valuable comments:
The "godfathers" of
#causalinference
in
#economics
, who triggered and fueled the so-called "credibility revolution" (finding and isolating random variation in data to credibly assess causal effects), receive their well-deserved prize. Congratulations!!!
#DataScience
BREAKING NEWS:
The 2021 Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel has been awarded with one half to David Card and the other half jointly to Joshua D. Angrist and Guido W. Imbens.
#NobelPrize
Two strands of research on
#Causal
#AI
&
#CausalMachineLearning
are: learning control variables when evaluating a predefined treatment; learning general causal structures, e.g. which treatments affect an outcome. Finally, the two strands are converging:
🔥Our study (joint with M Bia & L Lafférs) on treatment evaluation under sample selection (when a nonrandom subset of outcomes is unobserved) using double
#MachineLearning
has been published in the Journal of Business & Economic Statistics:
#EconTwitter
🚀 Check out our new working paper (joint with
@EconKevin
&
@lukaslaffers
) on testing the identification of causal effects in mediation & dynamic treatment models in observational data, with an application to Slovak labor market data:
#EconTwitter
Our paper (joint with Michela Bia &
@lukaslaffers
) on double machine learning for treatment evaluation with nonrandomly missing outcome variables has been published in the Journal of Business & Economic Statistics
#EconTwitter
#CausalTwitter
#StatsTwitter
🚨 Call for papers:
Workshop on
‘Machine Learning in Program Evaluation, High-dimensionality and Visualization Techniques’
June 20-21, 2023 in Belval, Luxembourg
Keynote speakers:
G. Imbens (
@guido_imbens
),
M. Huber (
@CausalHuber
),
S. Vansteelandt (
@SVansteelandt
)
1/2
🔥Our paper (joint with H Bodory, L Camponovo,
@MichaelLechner8
) on nonparametric bootstrap methods for propensity score matching (with a fixed number of matches) has been published in Statistics & Probability Letters:
#Econtwitter
#Causaltwitter
How economics has changed through the decades since the 1950s, as seen in the paper titles of the economics journals AER, JPE and QJE.
Paper by
@hugomoises
and Brice:
Ht
@DurRobert
🔥Working paper on estimating direct & indirect quantile treatment effects using double
#MachineLearning
: assessing
#CausalMechanisms
across ranks of the outcome distribution, e.g. different wage quantiles, while controlling for high-dimensional covariates:
How to distinguish the causal effect of a policy (eg unemployment benefits) on targeted individuals from spillover or general equilibrium effects in a region/market? This study outlines challenges and approaches:
#EconTwitter
#economics
#causalinference
🚨 New working paper🚨
Been working on this for a while, so I'm really excited to share this. I propose a new estimator combining the strength of existing research designs, along w/ a Stata command. This is still very much in developing stage so feedback is welcome!
A great article by
@MatteoCourthoud
on the intuition and implementation of causal trees for the data-driven detection of heterogeneous causal effects based on
#machinelearning
:
In beautiful
#T
übingen for giving a talk on
#causalinference
in the economics seminar of
@uni_tue
- thanks a lot, Martin Biewen and colleagues for hosting me 😀...and more to come in tomorrow's talk by
@MichaelLechner8
in the management seminar 😉
#datascience
I adapted the code of chapter 3 of
@CausalHuber
's book to Jupyter notebook to make it more accessible for the Python community.
I introduced some minor improvements and added some visuals.
You can check it here:
Ever wondered whether difference-in-differences estimation "works" for a multivalued/continuous (rather than binary) treatment, as e.g. the intensity of public good provision (such as childcare)? This paper discusses the required assumptions:
@LuigiBiagini
@yudapearl
@causalinf
Strictly speaking neither necessary nor sufficient. Confounding may imply a correlation that is not causal. And a non-monotonic causal relation may imply a zero correlation (even though this is a special case)
Join us for our Master program in
#DataAnalytics
&
#Economics
at
@ses_unifr
@unifr
, the first in Switzerland to integrate data analytics & AI with the economics of firms & markets, tailored to meet the demands of digitisation.
Recently accepted by EJ: "Time Preferences across Language Groups: Evidence on Intertemporal Choices from the Swiss Language Border" Holger Herz, Martin Huber, Tjaša Maillard-Bjedov & Svitlana Tyahlo
Just typed "Write a paper on instrumental variable regression consisting of 6000 words" into
#OpenAI
and this is what I got in a few minutes. Thinking about which journal I should submit it to...probably a lower-ranked one. Does the "Journal of AI-generated papers" already exist?