Research fields
  •  
Visa Ämnen

Estimation of causal effects for multivalued treatment

We consider a situation where the average treatment effect among the treatment (ATT) is a parameter of interest and where the treatment is multi-valued. The parameter is of interest when evaluating the effects of labour market programmes, for example. The development of the methods in the project builds on a framework with “generalized propensity scores” which is an augmentation of the propensity scores for a binary treatment. We develop estimators of the ATT using an outcome regression or inverse probability weighting. It is also possible to construct a doubly robust estimator of the ATT, using weighted ordinary least squares, thus combining the outcome regression with the inverse probability weighting. The estimators are defined in the framework of a partial M-estimation and we derive analytical expressions of the estimator and the asymptotic variance. We assess the finite sample properties in a simulation study.


Published by:

Anahid Zakinian

Changed:

6/20/2019