Dear Colleagues,
The Section of Statistical Learning and Data Science is pleased to present its July webinar. Dr. Karmakar will give a talk on a new paradigm of causal inference that allows for unmeasured confounders. Dr. Karmakar is an expert in causal inference, design and analysis of observational studies, and applications of statistics to public policy, health, social and marketing sciences. Hope to see you there!
Title: A new paradigm for causal inference in the presence of unmeasured confounders by calibrating a resistant population's variance
Speakers: Dr. Bikram Karmakar, Department of Statistics, University of Florida
Date and Time: July 30, 2024, 1:00 to 2:30 pm Eastern Time
Registration Link: https://www.eventbrite.com/e/a-new-paradigm-for-causal-inference-in-presence-of-unmeasured-confounders-tickets-949425636637?aff=oddtdtcreator. Click or tap if you trust this link." data-auth="Verified" rel="noopener noreferrer" target="_blank" href="https://nam10.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.eventbrite.com%2Fe%2Fa-new-paradigm-for-causal-inference-in-presence-of-unmeasured-confounders-tickets-949425636637%3Faff%3Doddtdtcreator&data=05%7C02%7Czhihuasu%40stat.ufl.edu%7C2a7cf4ddfce64d48259f08dca851134e%7C0d4da0f84a314d76ace60a62331e1b84%7C0%7C0%7C638570313380192702%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=gK2zi7vFd91Ewt3SFdA9zh9julVySCNQK1AwUf8ffC8%3D&reserved=0" data-linkindex="7">ASA SLDS Webinar Registration Link [eventbrite.com]
Abstract: Observational studies are alternatives to randomized trials to study the effect of a treatment where the treatment is not randomly assigned to the study units. The assumption that all the confounders between the treatment and outcome have been measured is central to causal inference from observational studies. However, the statistical methodology and empirical literature show diverging attitudes toward this assumption. A significant section of current methodological development assumes no unmeasured confounders and aims to find fine-tuned methods to estimate the overall or more specific effects. In contrast, nearly all published observational studies discuss the possibility of bias in their inference due to unmeasured confounders. This gap may be reduced by removing the untestable `no unmeasured confounders' assumption while also not asking for the additional special structures of the study, e.g., negative control or exogenous variability.
This webinar will discuss the fundamentals of the causal inference framework. We will examine the assumption of no unmeasured confounders and discuss the identification and estimation of the causal effects.
After that, in a general set-up that allows unmeasured confounding, we show that the conditional treatment effect can be identified as one of two possible values. We require (a) a nondeterministic treatment assignment, (b) that all effect modifiers are measured, and (c) a resistant population that was not exposed to the treatment or, if exposed, is unaffected by the treatment. Assumptions (a) and (b) are mild and (b) can be relaxed. For (c), which is a new assumption, we show that a resistant population is often available in practice. We develop a large sample inference methodology and demonstrate our proposed method in a study of the effect of surface mining in central Appalachia on birth weight that finds a harmful effect.
Presenter: Dr. Bikram Karmakar is an Assistant Professor in the Statistics Department at University of Florida. Prof Karmakar teaches advanced courses in statistics and probability and Graduate courses in statistical methodology. His research focuses on methodologies for causal deduction primarily motivated by social and health science problems, supported partly by grants from the US National Science Foundation and the National Institute of Health.
------------------------------
Zhihua Su, PhD
Associate Professor
Department of Statistics
University of Florida
zhihuasu@stat.ufl.edu------------------------------