Merge pull request #39 from Merck/prepcran

Simplify bootstrap functions and calculations

NAMESPACE

@@ -1,7 +1,6 @@
 # Generated by roxygen2: do not edit by hand
 
 export(calc_allrmds)
-export(calc_allrmds_boot)
 export(calc_haz_psm)
 export(calc_likes)
 export(calc_rmd)

R/resmeans.R

@@ -422,6 +422,7 @@ calc_rmd_first <- function(ds, cuttime) {
 #' @param discrate Discount rate (% per year)
 #' @param rmdmethod can be "int" (default for full integral calculations) or "disc" for approximate discretized calculations
 #' @param timestep required if method=="int", default being 1
+#' @param boot logical flag to indicate whether abbreviated output is required (default = FALSE), for example for bootstrapping
 #' @return List of detailed numeric results
 #' - cutadj indicates the survival function and area under the curves for PFS and OS up to the cutpoint
 #' - results provides results of the restricted means calculations, by model and state.
@@ -430,7 +431,7 @@ calc_rmd_first <- function(ds, cuttime) {
 #' \donttest{
 #' # Create dataset and fit survival models (splines)
 #' bosonc <- create_dummydata("flexbosms")
-#' fits <- fit_ends_mods_spl(bosonc)
+#' fits <- fit_ends_mods_par(bosonc)
 #' # Pick out best distribution according to min AIC
 #' params <- list(
 #'   ppd = find_bestfit(fits$ppd, "aic")$fit,
@@ -443,7 +444,7 @@ calc_rmd_first <- function(ds, cuttime) {
 #' # RMD using default "int" method, no lifetable constraint
 #' calc_allrmds(bosonc, dpam=params)
 #' # RMD using discretized ("disc") method, no lifetable constraint
-#' calc_allrmds(bosonc, dpam=params, rmdmethod="disc", timestep=1)
+#' calc_allrmds(bosonc, dpam=params, rmdmethod="disc", timestep=1, boot=TRUE)
 #' }
 calc_allrmds <- function(ptdata,
                          inclset = 0,
@@ -454,9 +455,8 @@ calc_allrmds <- function(ptdata,
                          lifetable = NA,
                          discrate = 0,
                          rmdmethod = "int",
-                         timestep = 1) {
-  # Set-up variables
-
+                         timestep = 1,
+                         boot = FALSE) {
   # Check calculations valid
   chvalid <- is.na(dpam[1])==FALSE
   if (chvalid==FALSE) stop("No validly fitted endpoints")
@@ -508,43 +508,9 @@ calc_allrmds <- function(ptdata,
   rmdres$pf <- rmdres$pf + first$pfarea
   rmdres$pd <- rmdres$pd + first$osarea - first$pfarea
   rmdres$os <- rmdres$os + first$osarea
-  return(list(cutadj=first, results=rmdres))
-}
-
-#' Wrapper to enable bootstrap sampling of restricted mean durations for each health state and all three models.
-#' @description Wrapper function to [calc_allrmds] to enable bootstrap sampling of calculations of restricted mean durations for each health state (progression free and progressed disease) for all three models (partitioned survival, clock forward state transition model, clock reset state transition model).
-#' @inheritParams calc_allrmds
-#' @return Numeric vector of restricted mean durations - PF for each model (PSM, STM-CF, STM-CR), then PD, then OS.
-#' @export
-#' @examples
-#' \donttest{
-#' bosonc <- create_dummydata("flexbosms")
-#' fits <- fit_ends_mods_spl(bosonc)
-#' # Pick out best distribution according to min AIC
-#' params <- list(
-#'   ppd = find_bestfit(fits$ppd, "aic")$fit,
-#'   ttp = find_bestfit(fits$ttp, "aic")$fit,
-#'   pfs = find_bestfit(fits$pfs, "aic")$fit,
-#'   os = find_bestfit(fits$os, "aic")$fit,
-#'   pps_cf = find_bestfit(fits$pps_cf, "aic")$fit,
-#'   pps_cr = find_bestfit(fits$pps_cr, "aic")$fit
-#' )
-#' calc_allrmds_boot(ptdata=bosonc, dpam=params)
-#' }
-calc_allrmds_boot <- function(ptdata,
-                              inclset = 0,
-                              dpam,
-                              cuttime = 0,
-                              Ty = 10,
-                              lifetable = NA,
-                              discrate = 0) {
-  if (inclset[1]==0) {inclset <- 1:length(ptdata$ptid)}
-  mb <- calc_allrmds(ptdata = ptdata,
-                     inclset = inclset,
-                     dpam = dpam,
-                     cuttime = cuttime,
-                     Ty = Ty,
-                     lifetable = lifetable,
-                     discrate = discrate)
-  c(mb$results$pf, mb$results$pd, mb$results$os)
+  if (boot==TRUE) {
+    c(rmdres$pf, rmdres$pd, rmdres$os)
+  } else {
+      list(cutadj=first, results=rmdres)
+  }
 }

vignettes/example.Rmd

@@ -182,17 +182,17 @@ The two STMs estimate a duration in the PF state slightly longer than the PSM. T
 
 The above output can be bootstrapped to generate standard errors. Here we use just 10 boostrap samples (`R=10`) just to illustrate the process. In practice, we would want to use far more than 10 samples.
 
-```{r boot, include=FALSE}
-# Bootstrap to calculate SE over R bootstrap samples
-bootlys <- boot(
+```{r boot}
+# Bootstrap to calculate SE over 10 bootstrap samples
+boot::boot(
   data = bosonc,
-  statistic = calc_allrmds_boot,
-  R = 10, # Number of simulations
+  statistic = calc_allrmds,
+  R = 10, # Number of samples
   cuttime = 0,
   Ty = 10,
-  dpam = params
+  dpam = params,
+  boot = TRUE
 )
-bootlys
 ```
 
 Note that the percentiles information reported indicates that in a small number of samples, the restricted mean duration in PD was restricted to be negative in the PSM. This indicates an inconsistency between the statistical models used in this case for modeling PFS and OS, and may be an additional reason why STMs may be preferred in this case.