日時：2026年4月3日(金) 13:30 ~　90分

講演者：Haruka Yoshida 先生（Yokohama National University）　

場所：webのみ　（※ 対象は特に限定しない）

題目： The Evaluation of Causal Effects in the Presence of Uncontrolled Confounding Bias

概要：
In standard statistical causal inference, reliable
evaluation of causal effects requires observing a sufficient set of
confounders. However, such information is often unavailable because of
technical and cost constraints, and uncontrolled confounding bias
remains one of the fundamental problems in causal inference.
In this talk, to solve the problem, I introduce two different
approaches for evaluating causal effects in the presence of
uncontrolled confounding bias.
First, I propose a Proportion-based Sensitivity Analysis (PSA) for
uncontrolled confounding bias. The key idea is that the mean squared
error (MSE) of the outcome variable with respect to the average causal
risk can be described as the sum of “the conditional variance of the
outcome variable given the exposure variable” and “the square of the
uncontrolled confounding bias”. In PSA, the sensitivity parameter is
formulated as the proportion of “the square of the uncontrolled
confounding bias” relative to the MSE, which enables practical
interpretation.
Second, I introduce Multiple Effect Restoration (MER), a causal
inference technique based on proxy variables, which is applicable when
key variables, including the exposure, the outcome, and confounders,
are observed with measurement errors. MER restores causal effects
using proxy variables for unobserved key variables and provides a
unified and broader class of identification conditions that includes
Effect Restoration and Proximal Causal Learning.

お問い合わせは以下のアドレスにメールでご連絡ください。
dser-center@biwako.shiga-u.ac.jp