Provides parameter estimation as well as variable selection in Finite Mixture of Accelerated Failure Time Regression and Finite Mixture of Regression Models. Furthermore, this package provides Ridge Regression and Elastic Net.
CHANGES SINCE LAST TIME o The argument 'activeset' is added which allow flexiblity in defining customized fmrs models in terms of which covariates and intercepts should be included in each component. o A bug regarding calculating residuals and fitted values is fixed. o An alternative code is added and will be activated when the results of MLE of weibull distribution create 'NaN'. It is based on a component-wise algorithm and using survreg function in survival package to find MLE.
BUG FIXES o Fixed a bug related to calculating fitted values and residuals.
CHANGES SINCE LAST TIME o We have implemented the Zhang(2010)'s suggestion for selecting extra tuning parameter in MCP penalty. o We have added an option for choosing extra tuning parameter in SICA penalty.
IMPROVEMENTS SINCE LAST TIME o We have implemented S4 classes and methods in this release.
CHANGES SINCE LAST TIME o Since S4 methods are implemented we have changed some arguments and definitions in this release. o fmrs object is changed to fmrsfit. o tunpar object is changed to fmrstunpar o You can no longer use $ to call the arguments of objects since they are written in S4 classes. Use @ instead. o New methods added to the package (see the manual).
IMPROVEMENTS SINCE LAST TIME o We have refined some of the codes to speed up the package.
FUTURE DEVELOPMENT o We intend to make some refinements to the package by introducing better S4 classes and methods. o We will add more distributions to the package including Generalized Gamma and Gompertz. o Since we are working on length-biased models, these models will be added to the package.
BUG FIXES o None reported so far.
IMPROVEMENTS SINCE LAST TIME o Compatibility with bioconductor is added.