Inference on Treatment Effects after Selection among High-Dimensional Controls

Belloni, Alexandre; Chernozhukov, Victor; Hansen, Christian

doi:10.1093/restud/rdt044

Cited by 1,234 publications

(1,178 citation statements)

References 41 publications

Supporting

Mentioning

1,167

Contrasting

Unclassified

Order By: Relevance

“…Second, when applying causal inference models to analyzing big data, there are high-dimensional econometric and machine learning techniques, such as LASSO (least absolute shrinkage and selection operator), the post-double-selection method, random forest, and bagging (bootstrap aggregating), that researchers can use to handle large data sets. Interested readers can refer to Tibshirani (1996), Belloni et al (2013Belloni et al ( , 2014, Varian (2014), Athey and Imbens (2017), and Wager and Athey (2017) for discussions on such methods. These methods have yet to see widespread applications.…”

Section: Resultsmentioning

confidence: 99%

OM Forum—Causal Inference Models in Operations Management

Lim

Reza

et al. 2017

M&SOM

View full text Add to dashboard Cite

Abstract. Operations management (OM) researchers have traditionally focused on developing normative mathematical models that prescribe what managers and firms should do. Recently, there has been increased interest in understanding what managers and firms actually do and the factors that drive these decisions. To advance this understanding, empirical investigation using causal inference models is critical. However, in many contexts, the ability to obtain causal inferences is fraught with the challenges of endogeneity and selection bias. This paper describes five empirical tools that have been widely used in economics to address these challenges and how they can be adopted by OM researchers. We also present an example that illustrates how the various attributes of big data-variety, velocity, and volume-can be useful in addressing the endogeneity bias.

show abstract

Section: Resultsmentioning

confidence: 99%

OM Forum—Causal Inference Models in Operations Management

Lim

Reza

et al. 2017

M&SOM

View full text Add to dashboard Cite

show abstract

“…Here, we report differences for all variables that are potentially included. Recall that only the subgroup of variables chosen by the double selection procedure (Belloni et al, 2014) is used in the propensity score estimations. See Table A1 educ abi 0XST 0educ abi 0Xmage90 0educ abi 0X f age40 0educ high voc 0X f age90 0educ high ed 0XAGREE 0 educ high ed 0Xmage40 0educ high ed 0Xmutter lebt 0educ missing 0XBE 0educ missing 0XSH 0 educ missing 0XRPSL 0 married 0XNRW 0divorced 0Xsah good 0divorced 0Xmage40 0widowed 0XRECI pos 0widowed 0X f age90 0 f oreign 0XNEURO 0 f oreign 0XSH 0 f oreign 0XHH 0 f oreign 0Xmage90 0 f oreign 0X f age90 0nkids hh 0Xsah good 0 nkids hh 0XNEURO 0 nkids hh 0XBW 0nkids hh 0Xmutter lebt 0single 0Xpartner 0single 0XCONSC 0single 0XHH 0single 0XBY 0 partner 0Xsah good 0 partner 0Xvater lebt 0partner 0Xmutter lebt 0age partn 0Xmutter lebt 0age partn 0X f ull time 0 sah good 0X f ull time 0sah satis 0Xmutter lebt 0sah vbad 0XHB 0sah vbad 0XBB 0doctor 0XRECI pos 0 NEURO 0Xmage80 0CONSC 0X f age60 0CONSC 0Xvater lebt 0CONSC 0Xmutter lebt 0CONSC 0X f ull time 0 AGREE 0X f age50 0OPENN 0Xmage50 0OPENN 0Xmage60 0EXTRA 0Xmage60 0RECI pos 0X f ull time 0 SH 0Xmage40 0HH 0Xy2012 0HH 0Xmage50 0HH 0Xmage60 0HB 0Xmage40 0HE 0X f age60 0RPSL 0Xmage90 0 BW 0X f ull time 0BB 0Xmage40 0ST 0Xmage90 0ST 0X f age40 0TH 0X f age90 0y2003 0Xmage40 0 y2003 0X f age30 0 y2004 0Xmage40 0y2006 0X f age40 0y2007 0X f ull time 0y2011 0Xmage40 0mage50 0X f age40 0 mage50 0Xmutter lebt 0 mage60 0X f age90 0mage60 0Xmutter lebt 0mage70 0X f age50 0mage70 0X f ull time 0mage80 0Xmutter lebt 0 Note: 0 indicate the time periods.…”

Section: Resultsmentioning

confidence: 99%

“…This procedure works if the "approximate sparsity assumption" (Belloni et al, 2014) holds which, stated verbally, implies the following: the chosen subset using the procedure described above leads to an approximation of the true relationship between outcome and controls, where the approximation error is sufficiently small. Thus, if the CIA holds given the full set of controls and their interactions, it also approximately holds for the chosen subset of controls.…”

Section: Control Variablesmentioning

confidence: 99%

Informal care and long-term labor market outcomes

Schmitz

Westphal

2017

Journal of Health Economics

View full text Add to dashboard Cite

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. The working papers published in the Series constitute work in progress circulated to stimulate discussion and critical comments. Views expressed represent exclusively the authors' own opinions and do not necessarily reflect those of the editors. Terms of use: Documents in Ruhr Economic Papers #646

show abstract

“…These algorithms do detect structure in y ˆ : when predictive quality is high, some structure must have been found. Some econometric results also show the converse: when there is structure, it will be recovered at least asymptotically (for example, for prediction consistency of LASSO-type estimators in an approximately sparse linear framework, see Belloni, Chernozhukov, and Hansen 2011). On the other hand, we have seen the dangers of naively interpreting the estimated β ˆ parameters as indicating the discovered structure.…”

Section: Recovering Structure: Estimation ( β ˆ ) Vs Prediction ( Y ˆ )mentioning

confidence: 91%

Machine Learning: An Applied Econometric Approach

Mullainathan

Spiess

2017

Journal of Economic Perspectives

1,354

803

View full text Add to dashboard Cite

Machines are increasingly doing “intelligent” things. Face recognition algorithms use a large dataset of photos labeled as having a face or not to estimate a function that predicts the presence y of a face from pixels x. This similarity to econometrics raises questions: How do these new empirical tools fit with what we know? As empirical economists, how can we use them? We present a way of thinking about machine learning that gives it its own place in the econometric toolbox. Machine learning not only provides new tools, it solves a different problem. Specifically, machine learning revolves around the problem of prediction, while many economic applications revolve around parameter estimation. So applying machine learning to economics requires finding relevant tasks. Machine learning algorithms are now technically easy to use: you can download convenient packages in R or Python. This also raises the risk that the algorithms are applied naively or their output is misinterpreted. We hope to make them conceptually easier to use by providing a crisper understanding of how these algorithms work, where they excel, and where they can stumble—and thus where they can be most usefully applied.

show abstract

Inference on Treatment Effects after Selection among High-Dimensional Controls

Cited by 1,234 publications

References 41 publications

OM Forum—Causal Inference Models in Operations Management

OM Forum—Causal Inference Models in Operations Management

Informal care and long-term labor market outcomes

Machine Learning: An Applied Econometric Approach

Contact Info

Product

Resources

About