Uncertainty in the estimation of the causal effect in observational studies is often due to unmeasured confounding, i.e., the presence of unobserved covariates linking treatments and outcomes. Instrumental Variables (IV) are commonly used to reduce the effects of unmeasured confounding. Existing methods for IV estimation either require strong parametric assumptions, use arbitrary distance metrics, or do not scale well to large datasets. We propose a matching framework for IV in the presence of observed categorical confounders that addresses these weaknesses. Our method first matches units exactly, and then consecutively drops variables to approximately match the remaining units on as many variables as possible. We show that our algorithm constructs better matches than other existing methods on simulated datasets, and we produce interesting results in an application to political canvassing.
机构:
Stanford Univ, Dept Hlth Res & Policy, Div Epidemiol, HRP Redwood Bldg, Stanford, CA 94305 USAStanford Univ, Dept Hlth Res & Policy, Div Epidemiol, HRP Redwood Bldg, Stanford, CA 94305 USA
机构:
NYU, Dept Biostat, New York, NY USANYU, Dept Biostat, New York, NY USA
Heng, Siyu
Zhang, Bo
论文数: 0引用数: 0
h-index: 0
机构:
Fred Hutchinson Canc Ctr, Vaccine & Infect Dis Div, Seattle, WA USANYU, Dept Biostat, New York, NY USA
Zhang, Bo
Han, Xu
论文数: 0引用数: 0
h-index: 0
机构:
Temple Univ, Dept Stat Operat & Data Sci, Philadelphia, PA USANYU, Dept Biostat, New York, NY USA
Han, Xu
Lorch, Scott A.
论文数: 0引用数: 0
h-index: 0
机构:
Childrens Hosp Philadelphia, Dept Pediat, Philadelphia, PA USANYU, Dept Biostat, New York, NY USA
Lorch, Scott A.
Small, Dylan S.
论文数: 0引用数: 0
h-index: 0
机构:
Univ Penn, Dept Stat & Data Sci, Philadelphia, PA 19104 USA
Univ Penn, Wharton Sch, Dept Stat & Data Sci, 265 South 37th St, Philadelphia, PA 19104 USANYU, Dept Biostat, New York, NY USA