**** 2011/10/18 - version 0.1 ****

This is the first version available for public download. We wait your comments for future improvements.

**** 2012/06/01 - version 0.2 ****

In this version, we have modified the title of the package in order to include the same methodology without censoring: the function boot.ROC() can be used when there is no censoring (logistic regression and traditional ROC curve), the function boot.ROCt() when there is censoring (Cox regression and time-dependent ROC curve). We also have modified the output values of the function: we added the results of the penalized regression model and the values of the false positive and negative rates for the different methods (apparent, cross-validation by bootstrap, bootstrap 0.632 and bootstrap 0.632+.

**** 2013/07/08 - version 0.3 ****

This version is compatible with R > 3.0.

**** 2013/12/26 - version 0.5 ****

This version is compatible with the CRAN policies. We therefore have changed the examples for time-saving. We also have merged the functions boot.ROC and boot.ROC.penfix into a single function boot.ROC. Similarly, we have merged the functions boot.ROCt and boot.ROCt.penfix into a single function boot.ROCt. The user have now to choose in the arguments of the function the option in order to define a priori the tuning parameter, which is by default re-estimated in each bootstrap sample. We also have performed many changes in the manual in order to offer a more precise support to users. This version is the first available on the CRAN mirrors.

**** 2014/10/22 - version 0.6 ****

We have corrected an error in the function boot.ROC. In the previous versions, when the tuning parameter was not defined by the user, the tuning parameter value used for each bootstrap was estimated only one time by cross-validation on the entire sample. This error is now corrected, the tuning parameter is estimated at each iteration when the user does not enter a fixed value.