arXiv subjects

Hong Xiong

Publications and source records attributed to Hong Xiong.

3 recordsLinked to original sources

arXiv2024

Comparative Effectiveness Research with Average Hazard for Censored Time-to-Event Outcomes: A Numerical Study

The average hazard (AH), recently introduced by Uno and Horiguchi, represents a novel summary metric of event time distributions, conceptualized as the general censoring-free average person-time incidence rate on a given time window, $[0,\tau].$ This metric is calculated as the ratio of the cumulative incidence probability at $\tau$ to the restricted mean survival time at $\tau$ and can be estimated through non-parametric methods. The AH's difference and ratio present viable alternatives to the traditional Cox's hazard ratio for quantifying the treatment effect on time-to-event outcomes in comparative clinical studies. While the methodology for evaluating the difference and ratio of AH in randomized clinical trials has been previously proposed, the application of the AH-based approach in general comparative effectiveness research (CER), where interventions are not randomly allocated, remains underdiscussed. This paper aims to introduce several approaches for applying the AH in general CER, thereby extending its utility beyond randomized trial settings to observational studies where treatment assignment is non-random.

stat.AP↗

arXiv2024

G-Transformer: Counterfactual Outcome Prediction under Dynamic and Time-varying Treatment Regimes

In the context of medical decision making, counterfactual prediction enables clinicians to predict treatment outcomes of interest under alternative courses of therapeutic actions given observed patient history. In this work, we present G-Transformer for counterfactual outcome prediction under dynamic and time-varying treatment strategies. Our approach leverages a Transformer architecture to capture complex, long-range dependencies in time-varying covariates while enabling g-computation, a causal inference method for estimating the effects of dynamic treatment regimes. Specifically, we use a Transformer-based encoder architecture to estimate the conditional distribution of relevant covariates given covariate and treatment history at each time point, then produces Monte Carlo estimates of counterfactual outcomes by simulating forward patient trajectories under treatment strategies of interest. We evaluate G-Transformer extensively using two simulated longitudinal datasets from mechanistic models, and a real-world sepsis ICU dataset from MIMIC-IV. G-Transformer outperforms both classical and state-of-the-art counterfactual prediction models in these settings. To the best of our knowledge, this is the first Transformer-based architecture that supports g-computation for counterfactual outcome prediction under dynamic and time-varying treatment strategies.

cs.LG↗

arXiv2020

Linear Regression Evaluation of Search Engine Automatic Search Performance Based on Hadoop and R

The automatic search performance of search engines has become an essential part of measuring the difference in user experience. An efficient automatic search system can significantly improve the performance of search engines and increase user traffic. Hadoop has strong data integration and analysis capabilities, while R has excellent statistical capabilities in linear regression. This article will propose a linear regression based on Hadoop and R to quantify the efficiency of the automatic retrieval system. We use R's functional properties to transform the user's search results upon linear correlations. In this way, the final output results have multiple display forms instead of web page preview interfaces. This article provides feasible solutions to the drawbacks of current search engine algorithms lacking once or twice search accuracies and multiple types of search results. We can conduct personalized regression analysis for user's needs with public datasets and optimize resources integration for most relevant information.

cs.IR↗