• Communications for Statistical Applications and Methods
• /
• v.21 no.4
• /
• pp.285-296
• /
• 2014

The quantile residual life function has been effectively used to interpret results from the analysis of the proportional hazards model for censored survival data; however, the quantile residual life function is not always estimable with currently available semi-parametric regression methods in the presence of heavy censoring. A parametric regression approach may circumvent the difficulty of heavy censoring, but parametric assumptions on a baseline hazard function can cause a potential bias. This article proposes a Bayesian semi-parametric regression approach for inference on an unknown baseline hazard function while adjusting for available covariates. We consider a model-based approach but the proposed method does not suffer from strong parametric assumptions, enjoying a closed-form specification of the parametric regression approach without sacrificing the flexibility of the semi-parametric regression approach. The proposed method is applied to simulated data and heavily censored survival data to estimate various quantile residual lifetimes and adjust for important prognostic factors.

• Communications for Statistical Applications and Methods
• /
• v.3 no.3
• /
• pp.113-123
• /
• 1996

Data-dependent (adaptive) choice of asymptotically efficient score functions for rank estimators and M-estimators of regression parameters in a linear regression model with left-truncated and right-censored data are developed herein. The locally adaptive smoothing techniques of Muller and Wang (1990) and Uzunogullari and Wang (1992) provide good estimates of the hazard function h and its derivative h' from left-truncated and right-censored data. However, since we need to estimate h'/h for the asymptotically optimal choice of score functions, the naive estimator, which is just a ratio of estimated h' and h, turns out to have a few drawbacks. An altermative method to overcome these shortcomings and also to speed up the algorithms is developed. In particular, we use a subroutine of the PPR (Projection Pursuit Regression) method coded by Friedman and Stuetzle (1981) to find the nonparametric derivative of log(h) for the problem of estimating h'/h.

• Communications for Statistical Applications and Methods
• /
• v.29 no.3
• /
• pp.333-351
• /
• 2022

The skew-t distribution is an attractive family of asymmetrical heavy-tailed densities that includes the normal, skew-normal and Student's-t distributions as special cases. In this work, we propose an EM-type algorithm for computing the maximum likelihood estimates for skew-t linear regression models with censored response. In contrast with previous proposals, this algorithm uses analytical expressions at the E-step, as opposed to Monte Carlo simulations. These expressions rely on formulas for the mean and variance of a truncated skew-t distribution, and can be computed using the R library MomTrunc. The standard errors, the prediction of unobserved values of the response and the log-likelihood function are obtained as a by-product. The proposed methodology is illustrated through the analyses of simulated and a real data application on Letter-Name Fluency test in Peruvian students.

• Journal of the Korean Statistical Society
• /
• v.28 no.2
• /
• pp.195-210
• /
• 1999

We consider a large class of exponential regression models with censored data and propose two modified Fisher scoring methods with corresponding algorithms. These proposed methods improve the Newton-Raphson method in estimating the model parameters. The simulated and real examples are illustrated in aspect of convergence.

• Communications for Statistical Applications and Methods
• /
• v.30 no.4
• /
• pp.389-402
• /
• 2023

Multivariate or clustered failure time data often occur in many medical, epidemiological, and socio-economic studies when survival data are collected from several research centers. If the data are periodically observed as in a longitudinal study, survival times are often subject to various types of interval-censoring, creating multivariate interval-censored data. Then, the event times of interest may be correlated among individuals who come from the same cluster. In this article, we propose a unified linear regression method for analyzing multivariate interval-censored data. We consider a semiparametric multivariate accelerated failure time model as a statistical analysis tool and develop a generalized Buckley-James method to make inferences by imputing interval-censored observations with their conditional mean values. Since the study population consists of several heterogeneous clusters, where the subjects in the same cluster may be related, we propose a generalized estimating equations approach to accommodate potential dependence in clusters. Our simulation results confirm that the proposed estimator is robust to misspecification of working covariance matrix and statistical efficiency can increase when the working covariance structure is close to the truth. The proposed method is applied to the dataset from a diabetic retinopathy study.

• Communications for Statistical Applications and Methods
• /
• v.15 no.2
• /
• pp.161-171
• /
• 2008

This study discusses Johnson's $S_U$-normal distribution capturing a wide range of non-normality in various regression models. We provide the likelihood inference using Johnson's $S_U$-normal distribution, and propose a likelihood ratio (LR) test for normality. We also apply the $S_U$-normal distribution to the binary and censored regression models. Monte Carlo simulations are used to show that the LR test using the $S_U$-normal distribution can be served as a model specification test for normal error distribution, and that the $S_U$-normal maximum likelihood (ML) estimators tend to yield more reliable marginal effect estimates in the binary and censored model when the error distributions are non-normal.

• Journal of Korean Society for Quality Management
• /
• v.27 no.2
• /
• pp.112-125
• /
• 1999

In a robust regression model, it is typically assumed that the errors are normally distributed. However, what if the error distribution is deviated from the normality and the response variables are not completely observable due to censoring? For complete data, Kim and Lai(1998) suggested a new adaptive M-estimator with an asymptotically efficient score function. The adaptive M-estimator is based on using B-splines to estimate the score function and simple cross validation to determine the knots of the B-splines, which are a modified version of Kun( 1992). We herein extend this method to right-censored data and study how well the adaptive M-estimator performs for various error distributions and censoring rates. Some impressive simulation results are shown.

• The Korean Journal of Applied Statistics
• /
• v.26 no.5
• /
• pp.737-746
• /
• 2013

The Bayesian method can be applied successfully to the estimation of the censored regression model introduced by Tobin (1958). The Bayes estimates show improvements over the maximum likelihood estimate; however, the performance of the Bayesian interval estimation is questionable. In Bayesian paradigm, the prior distribution usually reflects personal beliefs about the parameters. Such subjective priors will typically yield interval estimators with poor frequentist properties; however, an objective noninformative often yields a Bayesian procedure with good frequentist properties. We examine the performance of frequentist properties of noninformative priors for the Tobit regression model.

• The Korean Journal of Applied Statistics
• /
• v.29 no.7
• /
• pp.1311-1327
• /
• 2016

We propose a multi-state model for analyzing semi-competing risks data with interval-censored or missing intermediate events. This model is an extension of the 'illness-death model', which composes three states, such as 'healthy', 'diseased', and 'dead'. The state of 'diseased' can be considered as an intermediate event. Two more states are added into the illness-death model to describe missing events caused by a loss of follow-up before the end of the study. One of them is a state of 'LTF', representing a lost-to-follow-up, and the other is an unobservable state that represents the intermediate event experienced after LTF occurred. Given covariates, we employ the Cox proportional hazards model with a normal frailty and construct a full likelihood to estimate transition intensities between states in the multi-state model. Marginalization of the full likelihood is completed using the adaptive Gaussian quadrature, and the optimal solution of the regression parameters is achieved through the iterative Newton-Raphson algorithm. Simulation studies are carried out to investigate the finite-sample performance of the proposed estimation procedure in terms of the empirical coverage probability of the true regression parameter. Our proposed method is also illustrated with the dataset adapted from Helmer et al. (2001).

• The Korean Journal of Applied Statistics
• /
• v.35 no.3
• /
• pp.357-370
• /
• 2022

In this paper, Tobit and Heckit models are introduced. These models have been used for analyzing censored data. Censoring occurs at a specific point and a large number of observations are distributed with a positive probability at a certain point. Censoring can occur due to observing limitation or exogenous variables. Tobit and Heckit models are used to correct sample selection bias, which can occur when an ordinary linear regression model is fitted to censored data. However, the difference between the two models is not clearly accounted for; hence, they have often been used interchangeably. Therefore, the suitability of the models was validated through simulated data, and demonstrated through real data. As the result, it was confirmed that both Tobit and Heckit models are well-fitted to the data censored due to observing limitation, although Tobit model was fitted parsimoniously. In contrast, only Heckit model is well-fitted to the data censored due to exogenous variables.