10/n Our work suggests that using the debiased Ridge/Lasso in place of Ridge/Lasso can improve ATE estimation in high-dimensions, and emphasizes the importance of selecting the right methods for nuisance estimation. https://t.co/Lo9ZdFe5J3
1/n Excited to present my recent work with Martin Wainwright, “Challenges of the inconsistency regime: Novel debiasing methods for missing data models,” at tomorrow’s Online Causal Inference Seminar at 8:30am PT / 11:30 am ET on October 24! Thread below 👇 https://t.co/SwtRqV86my
9/n In our case, debiasing the outcome model requires estimation of the propensity model, even though neither is estimated consistently. Like AIPW, our method combines outcome and propensity model estimation, but it does so in a non-standard way.
Michael Celentano (MF 2021-2024) says that the Miller Fellowship "gives me flexibility & freedom in my research within a timeframe that allows me to undertake ambitious projects & explore new research directions." Nominate a Miller Fellow by 9/12/22: https://t.co/9utjYyqf0a
#STEM
Excited to present our recent work on exact asymptotics for the lasso with correlated features at the ISSI seminar tomorrow! Joint work with Yuting Wei and @Andrea__M. Seminar details here: https://t.co/ySjv30mOhi
Excited to share work with @Andrea__M and Yuting Wei on exact asymptotics for the lasso with correlated features at the Bernoulli-IMS One World Symposium. We provide applications to debiasing and tuning lambda.
Talk at https://t.co/VIV1s4mJOk. Paper at https://t.co/YzkqZEI0Gz.
For many years, Stanford Statistics Department has offered a free consulting service to Stanford's research community. For Summer 2020, we will open this service to anybody around the world for COVID-related data https://t.co/RZCfXUffqR