Table 1 Assumptions and features Cox, AFT, Portnoy, Wang and Wang, Bottai and Zhang, Yang and De Backer methods.
models | Assumptions | Advantages | Disadvantages |
|---|---|---|---|
Cox method | Proportional hazard | No need to consider a specific probability distribution for the survival time; Can used in many types of survival model | The effect of the included covariates is multiplicative The complexity of the HR estimate interpretation |
AFT method | The effect of a covariate is to accelerate or decelerate the life course of a disease by some constant Needs homogeneous covariates effect | direct interpretation of covariate effects on event time Can used in many types of survival model | Error term follow a specific probability distribution Failing to capture heterogeneity of covariate effects |
Portnoy method | The model at lower quantiles are all linear (global-linearity) | The effect of covariates is not restricted to be constant No distributional assumptions about the regression error term | The 'global' linearity assumption |
Bottai and Zhang method | The residuals follow a asymmetric Laplace distribution require -linearity assumption | The effect of covariates is not restricted to be constant Correct coverage and shorter computation time | Error term follow a Laplace distribution |
Wang and Wang method | Require a locally linear quantile regression | Not require global-linearity assumption | Requires estimating the true distribution of the outcome variable |
Yang method | Operates under the assumption that all the quantile functions are identifiable | Can handle different forms of censoring the estimator can achieve significant efficiency gains over the existing methods | It runs a risk of finding estimates even for non-identifiable quantile functions |
De Backer method | Require a locally linear quantile regression | Consistency and asymptotic normality of estimator | Restrict to the estimation of the classical linear regression model |
