使用右設限存活資料在比例勝算模型之下的變數選擇

Abstract

本篇論文中，我們專注於比例勝算模型分析右設限存活資料，因為該模型能夠直接預測存活機率，並且在Cox比例風險模型的假設不成立時，尤其適用。我們為了處理不完整的右設限資料，採用了Chiang等人（2018）提出的方法進行資料插補。資料插補完後，在不同時間點下，我們將比例勝算模型作為羅吉斯迴歸模型進行估計，並應用了如LASSO、彈性網、自適應LASSO和自適應彈性網等正則化技術進行變數選擇與係數估計。我們最後提出了一個準則，這個準則提供了一個創新的選變量方法。

In this thesis, we concentrate on the proportional odds model for right-censored survival data, as it directly predicts survival probabilities and is particularly suitable when the assumptions of the Cox proportional hazards model are not met. To address incomplete right-censored data, we utilize the approach proposed by Chiang et al. (2018) to impute data. After imputation, we estimate the proportional odds model at various time points, treating it as a logistic regression model, and apply regularization techniques, including LASSO, Elastic-Net, adaptive LASSO and adaptive Elastic-Net for variable selection and coefficients estimation simultaneously. Finally, we propose a criterion that introduces a novel approach to variable selection.