Fay, Michael (NIH/NIAID) [E]
2016-Feb-01 14:01 UTC
[R-pkgs] Update on bpcp R package: Confidence intervals to use with Kaplan-Meier Survival Estimator
R Users, We have updated the bpcp R package. It gives pointwise confidence intervals for a survival distribution from right censored data. It is not based on asymptotic approximations, so it may be used with any sample size, and with any censoring distribution, as long as the censoring is non-informative. Extensive simulations show that the bpcp confidence intervals appear to guarantee coverage. When there is no censoring the method reduces to the exact binomial intervals for the proportion of survivors at each time point. Here are some important improvements: 1. The bpcp function now includes a midp option. This provides an confidence interval that is closer to the nominal level "on average" over the values of the parameter. It reduces to the mid-p confidence interval for a binomial parameter when there is no censoring. Just like other mid-p confidence intevals, for some values of the parameter it can be conservative, and for some values it can be slightly anti-conservative. With extensive censoring it can be very conservative, because (as with the original version) no assumptions are made about the censoring distribution. 2. There is a new convention for the original bpcp exactly at the failure times. This new convention ensures that the bpcp confidence intervals enclose the Kaplan-Meier estimator. 3. There is a new option to enforce monotonicity. This is rarely needed, but is used by default. All new simulations were done using this enforced monotonicity. 4. An error was fixed in the method that only occurred when non-default Delta values were used. References: Fay, MP, Brittain, EH (2016). Finite Sample Pointwise Confidence Intervals for a Survival Distribution with Right-Censored Data. (to appear in Statistics in Medicine, outlines midp modification, and provides many more simulations). Fay, MP, Brittain, EH, and Proschan, MA (2013). Pointwise confidence intervals for a survival distribution with small samples or heavy censoring. Biostatistics. 14(4): 723-736. Thanks, Let me know if you have comments or want a preprint of the 2016 paper. Mike ************************************** Michael P. Fay, PhD Mathematical Statistician Biostatistics Research Branch/DCR/NIAID 5601 Fishers Lane, Room 4B53 Rockville, MD 20852 240-669-5228 For FexEx/UPS use: Rockville, MD 20852 For US Mail use: Bethesda, MD 20892 Disclaimer: The information in this e-mail and any of its attachments is confidential and may contain sensitive information. It should not be used by anyone who is not the original intended recipient. If you have received this e-mail in error please inform the sender and delete it from your mailbox or any other storage devices. The National Institute of Allergy and Infectious Diseases (NIAID) shall not accept liability for any statement made that are the sender's own and not expressly made on behalf of the NIAID by one of its representatives. [[alternative HTML version deleted]]