#  Naijia Liu (Applied Stats Workshop) 

 



####  calendar\_today Date and Time 

 **April 16, 2025** 

 12:00PM - 01:30PM EDT 

####  pin\_drop Location 

 **CGIS Knafel room K354, or virtual via Zoom**  



 

 



 

## Today's Speaker

Naijia Liu, "Missing Data in Confounders" (w/Melody Huang)

## Abstract

When confounder data is missing-not-at-random in observational data, standard assumptions are no longer sufficient for identifying the average treatment effect. In practice, researchers often rely on either a complete case estimator, or imputation to estimate the average treatment effect. In the following paper, we demonstrate that despite their popularity, neither approach provides unbiased estimation of the ATE, without additional assumptions that are untenable in practice. We show that the imputation estimators will only be unbiased in settings when we are able to perfectly impute the missing confounder values. Paradoxically, this is only feasible in settings when the missing values can be perfectly explained by the observed data, making it redundant to impute at all. We propose an alternative identification strategy, which allows researchers to leverage a two-stage estimator to unbiasedly estimate the ATE under arguably weaker assumptions. We introduce a suite of validation approaches to evaluate the credibility of the proposed assumptions. We illustrate our proposed framework on a recent study evaluating the impact of government transparency on state legislature and show that results from the proposed two-stage estimator differ significantly from existing missing data estimators

*The Applied Statistics Workshop (Gov 3009) meets all academic year, Wednesdays, 12pm-1:30pm, in CGIS K354. This workshop is a forum for advanced graduate students, faculty, and visiting scholars to present and discuss methodological or empirical work in progress in an interdisciplinary setting. The workshop features a tour of Harvard's statistical innovations and applications with weekly stops in different fields and disciplines and includes occasional presentations by invited speakers.*  
  
*All interested Harvard affiliates are invited to attend.*



 

 



 

 See also:- [ Workshops, Seminars & Classes ](/event-type/workshops-seminars-classes)
 
 

 Share on:- [     Facebook ](#)
- [     Twitter ](#)
- [     Linkedin ](#)
 


 Save: [ Add to calendar calendar\_today ](https://www.iq.harvard.edu/node/1940936/event-feed.ics)  Copy link link