The application of Firth’s procedure to Cox and logistic regression (original) (raw)
Related papers
A solution to the problem of monotone likelihood in Cox regression
2001
Summary. The phenomenon of monotone likelihood is observed in the fitting process of a Cox model if the likelihood converges to a finite value while at least one parameter estimate diverges to±∞. Monotone likelihood primarily occurs in small samples with substantial censoring of survival times and several highly predictive covariates. Previous options to deal with monotone likelihood have been unsatisfactory.
FC06: A SAS® macro for Cox regression with Firth’s penalization
2006
Abstract The phenomenon of monotone likelihood is observed in the fitting process of a Cox model if the likelihood converges while at least one parameter estimate diverges to±infinity. Monotone likelihood primarily occurs in small samples with several unbalanced and highly predictive covariates, and with a high percentage of censoring. A procedure by Firth (1993) originally developed to reduce the bias of maximum likelihood estimates is shown to provide an ideal solution to monotone likelihood (cf. Heinze & Schemper, 2001).
A solution to the problem of separation in logistic regression
Statistics in medicine, 2002
The phenomenon of separation or monotone likelihood is observed in the ÿtting process of a logistic model if the likelihood converges while at least one parameter estimate diverges to ± inÿnity. Separation primarily occurs in small samples with several unbalanced and highly predictive risk factors. A procedure by Firth originally developed to reduce the bias of maximum likelihood estimates is shown to provide an ideal solution to separation. It produces ÿnite parameter estimates by means of penalized maximum likelihood estimation. Corresponding Wald tests and conÿdence intervals are available but it is shown that penalized likelihood ratio tests and proÿle penalized likelihood conÿdence intervals are often preferable. The clear advantage of the procedure over previous options of analysis is impressively demonstrated by the statistical analysis of two cancer studies.
Asymptotic Methods in Probability and Statistics, 1998
Chapter 565 Cox Regression Introduction This procedure performs Cox (proportional hazards) regression analysis, which models the relationship between a set of one or more covariates and the hazard rate. Covariates may be discrete or continuous. Cox's proportional hazards regression model is solved using the method of marginal likelihood outlined in Kalbfleisch (1980). This routine can be used to study the impact of various factors on survival. You may be interested in the impact of diet, age, amount of exercise, and amount of sleep on the survival time after an individual has been diagnosed with a certain disease such as cancer. Under normal conditions, the obvious statistical tool to study the relationship between a response variable (survival time) and several explanatory variables would be multiple regression. Unfortunately, because of the special nature of survival data, multiple regression is not appropriate. Survival data usually contain censored data and the distribution of survival times is often highly skewed. These two problems invalidate the use of multiple regression. Many alternative regression methods have been suggested. The most popular method is the proportional hazard regression method developed by Cox (1972). Another method, Weibull regression, is available in NCSS in the Distribution Regression procedure. Further Reading Several books provide in depth coverage of Cox regression. These books assume a familiarity with basic statistical theory, especially with regression analysis. Collett (1994) provides a comprehensive introduction to the subject. Hosmer and Lemeshow (1999) is almost completely devoted to this subject. Therneau and Grambsch (2000) provide a complete and up-to-date discussion of this subject. We found their discussion of residual analysis very useful. Klein and Moeschberger (1997) provides a very readable account of survival analysis in general and includes a lucid account of Cox regression. The Cox Regression Model Survival analysis refers to the analysis of elapsed time. The response variable is the time between a time origin and an end point. The end point is either the occurrence of the event of interest, referred to as a death or failure, or the end of the subject's participation in the study. These elapsed times have two properties that invalidate standard statistical techniques, such as t-tests, analysis of variance, and multiple regression. First of all, the time values are often positively skewed. Standard statistical techniques require that the data be normally distributed. Although this skewness could be corrected with a transformation, it is easier to adopt a more realistic data distribution.
Full likelihood inferences in the Cox model: an empirical likelihood approach
Annals of the Institute of Statistical Mathematics, 2011
For the regression parameter β 0 in the Cox model, there have been several estimators constructed based on various types of approximated likelihood, but none of them has demonstrated small-sample advantage over Cox's partial likelihood estimator. In this article, we derive the full likelihood function for (β 0 , F 0 ), where F 0 is the baseline distribution in the Cox model. Using the empirical likelihood parameterization, we explicitly profile out nuisance parameter F 0 to obtain the full-profile likelihood function for β 0 and the maximum likelihood estimator (MLE) for (β 0 , F 0 ). The relation between the MLE and Cox's partial likelihood estimator for β 0 is made clear by showing that Taylor's expansion gives Cox's partial likelihood estimating function as the leading term of the full-profile likelihood estimating function. We show that the log fulllikelihood ratio has an asymptotic chi-squared distribution, while the simulation studies indicate that for small or moderate sample sizes, the MLE performs favorably over Cox's partial likelihood estimator. In a real dataset example, our full likelihood ratio test and Cox's partial likelihood ratio test lead to statistically different conclusions.
Malaysian Journal of Science, 2015
Traditional inferential procedures based on the asymptotic normality assumption such as the Wald often produce misleading inferences when dealing with censored data and small samples. Alternative estimation techniques such as the jackknife and bootstrap percentile allow us to construct the interval estimates without relying on any classical assumptions. Recently, the double bootstrap became preferable as it is not only free from any classical assumptions, but also has higher order of accuracy. In this paper, we compare the performances of the interval estimates based on the double bootstrap without pivot with the Wald, jackknife and bootstrap percentile interval estimates for the parameters of the log logistic model with right censored data and covariates via a coverage probability study. Based on the results of the study, we concluded that the double bootstrap without pivot technique works better than the other interval estimation techniques, even when sample size is 25. The double bootstrap without pivot technique worked well with real data on hypernephroma patients.
SAS and SPLUS programs to perform Cox regression without convergence problems
2002
When analyzing survival data, the parameter estimates and consequently the relative risk estimates of a Cox model sometimes do not converge to finite values. This phenomenon is due to special conditions in a data set and is known as 'monotone likelihood'. Statistical software packages for Cox regression using the maximum likelihood method cannot appropriately deal with this problem. A new procedure to solve the problem has been proposed by G. Heinze, M.
The Log-gamma-logistic Regression Model: Estimation, Sensibility and Residual Analysis
Journal of Statistical Theory and Applications, 2017
In this paper, we formulate and develop a log-linear model using a new distribution called the log-gammalogistic. We show that the new regression model can be applied to censored data since it represents a parametric family of models that includes as sub-models several widely-known regression models and therefore can be used more effectively in the analysis of survival data. We obtain maximum likelihood estimates of the model parameters by considering censored data and evaluate local influence on the estimates of the parameters by taking different perturbation schemes. Some global-influence measurements are also investigated. Further, for different parameter settings, sample sizes and censoring percentages, various simulations are performed. In addition, the empirical distributions of some modified residuals are displayed and compared with the standard normal distribution. These studies suggest that the residual analysis usually performed in normal linear regression models can be extended to modified deviance residuals in the proposed regression model applied to censored data. We demonstrate that our extended regression model is very useful to the analysis of real data and may give more realistic fits than other special regression models.
Robust and Efficient Estimation in the Parametric Cox Regression Model under Random Censoring
2018
Cox proportional hazard regression model is a popular tool to analyze the relationship between a censored lifetime variable with other relevant factors. The semi-parametric Cox model is widely used to study different types of data arising from applied disciplines like medical science, biology, reliability studies and many more. A fully parametric version of the Cox regression model, if properly specified, can yield more efficient parameter estimates leading to better insight generation. However, the existing maximum likelihood approach of generating inference under the fully parametric Cox regression model is highly non-robust against data-contamination which restricts its practical usage. In this paper we develop a robust estimation procedure for the parametric Cox regression model based on the minimum density power divergence approach. The proposed minimum density power divergence estimator is seen to produce highly robust estimates under data contamination with only a slight loss...