• Journal of the Korean Data and Information Science Society
• /
• v.25 no.1
• /
• pp.211-218
• /
• 2014

Qu et. al. (2000) proposed the quadratic inference functions (QIF) method to marginal model analysis of longitudinal data to improve the generalized estimating equations (GEE). It yields a substantial improvement in efficiency for the estimators of regression parameters when the working correlation is misspecified. But for the longitudinal data with time-dependent covariates, when the implicit full covariates conditional mean (FCCM) assumption is violated, the QIF can not provide more consistent and efficient estimator than GEE (Cho and Dashnyam, 2013). Lai and Small (2007) divided time-dependent covariates into three types and proposed generalized method of moment (GMM) for longitudinal data with time-dependent covariates. They showed that their GMM type II and GMM moment selection methods can be more ecient than GEE with independence working correlation (GEE-ind) in the case of type II time-dependent covariates. We develop upgraded QIF method for type II time-dependent covariates. We show that this upgraded QIF method can provide substantial gains in efficiency over QIF and GEE-ind in the case of type II time-dependent covariates.

• Journal of the Korean Data and Information Science Society
• /
• v.24 no.4
• /
• pp.877-883
• /
• 2013

The quadratic inference functions (QIF) method proposed by Qu et al. (2000) and the generalized method of moments (GMM) for marginal regression analysis of longitudinal data with time-dependent covariates proposed by Lai and Small (2007) both are the methods based on generalized method of moment (GMM) introduced by Hansen (1982) and both use generalized estimating equations (GEE). Lai and Small (2007) divided time-dependent covariates into three types such as: Type I, Type II and Type III. In this paper, we compared these methods in the case of Type II and Type III in which full covariates conditional mean assumption (FCCM) is violated and interested in whether they can improve the results of GEE with independence working correlation. We show that in the marginal regression model with Type II time-dependent covariates, GMM Type II of Lai and Small (2007) provides more ecient result than QIF and for the Type III time-dependent covariates, QIF with independence working correlation and GMM Type III methods provide the same results. Our simulation study showed the same results.

• Journal of Gastric Cancer
• /
• v.15 no.4
• /
• pp.238-245
• /
• 2015

Purpose: This study aimed to identify time-dependent prognostic factors and demonstrate the time-dependent effects of important prognostic factors in patients with advanced gastric cancer (AGC). Materials and Methods: We retrospectively evaluated 3,653 patients with AGC who underwent curative standard gastrectomy between 1991 and 2005 at the Korea Cancer Center Hospital. Multivariate survival analysis with Cox proportional hazards regression was used in the analysis. A non-proportionality test based on the Schoenfeld residuals (also known as partial residuals) was performed, and scaled Schoenfeld residuals were plotted over time for each covariate. Results: The multivariate analysis revealed that sex, depth of invasion, metastatic lymph node (LN) ratio, tumor size, and chemotherapy were time-dependent covariates violating the proportional hazards assumption. The prognostic effects (i.e., log of hazard ratio [LHR]) of the time-dependent covariates changed over time during follow-up, and the effects generally diminished with low slope (e.g., depth of invasion and tumor size), with gentle slope (e.g., metastatic LN ratio), or with steep slope (e.g., chemotherapy). Meanwhile, the LHR functions of some covariates (e.g., sex) crossed the zero reference line from positive (i.e., bad prognosis) to negative (i.e., good prognosis). Conclusions: The time-dependent effects of the prognostic factors of AGC are clearly demonstrated in this study. We can suggest that time-dependent effects are not an uncommon phenomenon among prognostic factors of AGC.

• The Korean Journal of Applied Statistics
• /
• v.35 no.1
• /
• pp.35-47
• /
• 2022

The receiver operating characteristic (ROC) curve was developed to quantify the classification ability of marker values (covariates) on the response variable and has been extended to survival data with diverse missing data structure. When survival data is understood as binary data (status of being alive or dead) at each time point, the ROC curve expressed at every time point results in time-dependent ROC curve and time-dependent area under curve (AUC). In particular, a follow-up study brings the change of cohort and incomplete data structures such as censoring and competing risk. In this paper, we review time-dependent ROC estimators under several contexts and perform simulation to check the performance of each estimators. We analyzed a dementia dataset to compare the prognostic power of markers.

• Management Science and Financial Engineering
• /
• v.12 no.2
• /
• pp.19-35
• /
• 2006

We use the hierarchical Bayesian approach to describe the transition probabilities of a binary nonhomogeneous Markov chain. The Markov chain is used for describing the transition behavior of emotionally disturbed children in a treatment program. The effects of covariates on transition probabilities are assessed using a logit link function. To describe the time evolution of transition probabilities, we consider two modeling strategies. The first strategy is based on the concept of exchangeabiligy, whereas the second one is based on a first order Markov property. The deviance information criterion (DIC) measure is used to compare models with two different time dependent structures. The inferences are made using the Markov chain Monte Carlo technique. The developed methodology is applied to some real data.

• Journal of Korea Water Resources Association
• /
• v.42 no.6
• /
• pp.445-455
• /
• 2009

In this paper proportional hazards models for the first through seventh break of 150 mm cast iron pipes in a case study area are established. During the modeling process the assumption of the proportional hazards for covariates on the hazards is examined to include the time-dependent covariate terms in the models. As a result, the pipe material/joint type and the number of customers are modeled as time-dependent for the first failure, and for the second failure only the number of customers is modeled as time-dependent. From the analysis on the baseline hazard functions the failure hazards are found to be generally increasing for the first and second failure, while the hazards of the third break and beyond showed a form of a bath-tub. Furthermore, the changes in the baseline hazard rates according to the time and number of break reflect that the general condition of the pipes is deteriorating. The factors causing pipe break and their effects are analyzed based on the estimated regression coefficients and their hazard ratios, and the constructed models are verified using the deviance residuals of the models.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.2
• /
• pp.349-359
• /
• 2017

In this paper, we conducted survival analyses by fitting the Cox proportional hazards model to stage III proximal colon cancer data obtained from the Surveillance, Epidemiology, and End Results program of the National Cancer Institute. We investigated the effect of covariates on the hazard function for death from proximal colon cancer in stage III with surgery performed and estimated the survival probability for a patient with specific covariates. We showed that the proportional hazards assumption is satisfied for covariates that were used to analyses, using a test based on the Schoenfeld residuals and plots of the Schoenfeld residuals and $log[-log\{{\hat{S}}(t)\}]$. We evaluated the model calibration and discriminatory accuracy by calibration plot and time-dependent area under the ROC curve, which were calculated using 10-fold cross validation.

• The Korean Journal of Applied Statistics
• /
• v.21 no.2
• /
• pp.291-299
• /
• 2008

This paper discusses model building for repeated binary response data with different time-dependent covariates each occasion. Since repeated measurements data are having correlated structure, weighed least squares(WLS) methodology is applied. Repeated measures designs are usually having different sizes of experimental units like split-plot designs. However repeated measures designs differ from split-plot designs in that the levels of one or more factors cannot be randomly assigned to one or more of the sizes of experimental units in the experiment. In this case, the levels of time cannot be assigned at random to the time intervals. Because of this nonrandom assignment, the errors corresponding to the respective experimental units may have a covariance matrix. So, the estimates of effects included in a suggested logit model are obtained by using covariance structures.

• Journal of Preventive Medicine and Public Health
• /
• v.54 no.6
• /
• pp.395-403
• /
• 2021

Objectives: Previous studies have shown that participation in social activities (SA) can prevent cognitive decline (CD) and that living arrangements (LA) can affect cognitive function. This study aimed to evaluate the effects of SA and LA on CD, as well as their interactions, using longitudinal data. Methods: Data were used from the 2006-2018 Korean Longitudinal Study for Aging, which followed 10 254 adults older than 45 years over a 12-year period. CD was defined as a ≥4-point score decrease in the Mini-Mental Status Exam over 2 years. We developed an extended Cox proportional hazards model for time-dependent covariates to estimate the hazard ratio (HR) of CD in 4 groups: (1) socially active and living with others, (2) socially active and living alone, (3) socially inactive and living with others (SILO), and (4) socially inactive and living alone (SILA). The model was stratified by gender and adjusted for important confounders. Results: The HR of CD was significantly higher in the SILO group in men (HR,1.36; 95% confidence interval [CI], 1.08 to 1.78) and in the SILA group in women (HR, 1.72; 95% CI, 1.08 to 2.75). However, the interaction term for gender was not significant. Conclusions: Among socially inactive elderly adults, the HR of CD was elevated in men who lived with others and in women who lived alone, although the interaction term for gender was not significant. Socially inactive men who live with others and socially inactive women who live alone are particularly encouraged to participate in SA to prevent CD.

• Genomics & Informatics
• /
• v.20 no.2
• /
• pp.23.1-23.9
• /
• 2022

A survival prediction model has recently been developed to evaluate the prognosis of resected nonmetastatic pancreatic ductal adenocarcinoma based on a Cox model using two nationwide databases: Surveillance, Epidemiology and End Results (SEER) and Korea Tumor Registry System-Biliary Pancreas (KOTUS-BP). In this study, we applied two machine learning methods-random survival forests (RSF) and support vector machines (SVM)-for survival analysis and compared their prediction performance using the SEER and KOTUS-BP datasets. Three schemes were used for model development and evaluation. First, we utilized data from SEER for model development and used data from KOTUS-BP for external evaluation. Second, these two datasets were swapped by taking data from KOTUS-BP for model development and data from SEER for external evaluation. Finally, we mixed these two datasets half and half and utilized the mixed datasets for model development and validation. We used 9,624 patients from SEER and 3,281 patients from KOTUS-BP to construct a prediction model with seven covariates: age, sex, histologic differentiation, adjuvant treatment, resection margin status, and the American Joint Committee on Cancer 8th edition T-stage and N-stage. Comparing the three schemes, the performance of the Cox model, RSF, and SVM was better when using the mixed datasets than when using the unmixed datasets. When using the mixed datasets, the C-index, 1-year, 2-year, and 3-year time-dependent areas under the curve for the Cox model were 0.644, 0.698, 0.680, and 0.687, respectively. The Cox model performed slightly better than RSF and SVM.