arXiv subjects

Takashi Sozu

Publications and source records attributed to Takashi Sozu.

3 recordsLinked to original sources

arXiv2025

A comparative study of augmented inverse propensity weighted estimators using outcome-oriented covariate selection via penalization with outcome-adaptive lasso

When estimating causal effects from observational data with numerous covariates, employing penalized covariate selection can improve the estimation efficiency. Outcome-oriented covariate selection, which involves selecting covariates related to the outcome, can enhance efficiency, even for propensity score (PS) methods. For outcome-oriented covariate selection in PS models, outcome-adaptive lasso (OAL) can be used for penalization with the oracle property. The performance of inverse propensity weighted (IPW) estimators using the OAL was shown to be superior to that of the IPW estimators using other covariate selection methods for parametric models. However, the augmented IPW (AIPW) estimator is typically employed as a doubly robust estimator for the average treatment effect, which requires both PS and outcome models. Despite this, which covariate selection method for outcome models should be combined with the OAL to form the AIPW estimator remains unclear. We evaluated the performance of the AIPW estimators using the OAL for PS models and various outcome-oriented covariate selection via penalization for outcome models. We conducted numerical experiments to evaluate the performance of AIPW estimators using various covariate selection via penalization. The performance of the AIPW estimators using outcome-oriented covariate selection via penalization with the oracle property for both PS and outcome models was superior to that of the other estimators and similar to that of the AIPW estimator, which relies on true confounders and outcome predictors. In contrast, the bias of the AIPW estimators not relying on the oracle property was high. In a clinical trial dataset analysis, the AIPW estimators using outcome-oriented covariate selection via penalization with and without the oracle property showed similar estimates and standard errors.

stat.ME↗

arXiv2024

Evaluating marginal likelihood approximations of dose-response relationship models in Bayesian benchmark dose methods for risk assessment

Benchmark dose (BMD; a dose associated with a specified change in response) is used to determine the point of departure for the acceptable daily intake of substances for humans. Multiple dose-response relationship models are considered in the BMD method. Bayesian model averaging (BMA) is commonly used, where several models are averaged based on their posterior probabilities, which are determined by calculating the marginal likelihood (ML). Several ML approximation methods are employed in standard software packages, such as BBMD, \texttt{ToxicR}, and Bayesian BMD for the BMD method, because the ML cannot be analytically calculated. Although ML values differ among approximation methods, resulting in different posterior probabilities and BMD estimates, this phenomenon is neither widely recognized nor quantitatively evaluated. In this study, we evaluated the performance of five ML approximation methods: (1) maximum likelihood estimation (MLE)-based Schwarz criterion, (2) Markov chain Monte Carlo (MCMC)-based Schwarz criterion, (3) Laplace approximation, (4) density estimation, and (5) bridge sampling through numerical examples using four real experimental datasets. Eight models and three prior distributions used in BBMD and \texttt{ToxicR} were assumed. The approximation and estimation biases of bridge sampling were the smallest regardless of the dataset or prior distributions. Both the approximation and estimation biases of MCMC-based Schwarz criterion and Laplace approximation were large for some datasets. Thus, the approximation biases of the density estimation were relatively small but were large for some datasets. In terms of the accuracy of ML approximation methods, using Bayesian BMD, in which the bridge sampling is available, is preferred.

stat.CO↗

arXiv2024

Nonparametric Bayesian approach for dynamic borrowing of historical control data

When incorporating historical control data into the analysis of current randomized controlled trial data, it is critical to account for differences between the datasets. When the cause of the difference is an unmeasured factor and adjustment for observed covariates only is insufficient, it is desirable to use a dynamic borrowing method that reduces the impact of heterogeneous historical controls. We propose a nonparametric Bayesian approach for borrowing historical controls that are homogeneous with the current control. Additionally, to emphasize the resolution of conflicts between the historical controls and current control, we introduce a method based on the dependent Dirichlet process mixture. The proposed methods can be implemented using the same procedure, regardless of whether the outcome data comprise aggregated study-level data or individual participant data. We also develop a novel index of similarity between the historical and current control data, based on the posterior distribution of the parameter of interest. We conduct a simulation study and analyze clinical trial examples to evaluate the performance of the proposed methods compared to existing methods. The proposed method based on the dependent Dirichlet process mixture can more accurately borrow from homogeneous historical controls while reducing the impact of heterogeneous historical controls compared to the typical Dirichlet process mixture. The proposed methods outperform existing methods in scenarios with heterogeneous historical controls, in which the meta-analytic approach is ineffective.

stat.ME↗