correct codes

R/maxcombo.R

@@ -110,7 +110,8 @@ maxcombo <- function(data = sim_pw_surv(n = 200) |> cut_data_by_event(150),
   # Tidy outputs ----
   ans <- list()
   ans$method <- "Maxcombo"
-  temp <- data.frame(rho = rho, gamma = gamma) %>% mutate(x= paste0("FH(", rho, ", ", gamma, ")"))
+  temp <- data.frame(rho = rho, gamma = gamma)
+  temp$x <- paste0("FH(", temp$rho, ", ", temp$gamma, ")")
   ans$parameter <- paste(temp$x, collapse = " + ")
   ans$estimation <- NULL
   ans$se <- NULL

R/wlr.R

@@ -51,31 +51,25 @@ wlr <- function(data, weight, return_variance = FALSE) {
   if (inherits(weight, "fh")) {
     x <- x |> fh_weight(rho = weight$rho, gamma = weight$gamma)
 
-    x$weighted_o_minus_e <- x$weight * x$o_minus_e
-    x$weighted_var <- x$weight^2 * x$var_o_minus_e
-
-    weighted_var_total <- sum(x$weighted_var)
-    weighted_o_minus_e_total <- sum(x$weighted_o_minus_e)
-
     ans$parameter <- paste0("FH(rho=", weight$rho, ", gamma =", weight$gamma, ")")
-    ans$estimation <- weighted_o_minus_e_total
-    ans$se <- sqrt(weighted_var_total)
+    ans$estimation <- sum(x$weight * x$o_minus_e)
+    ans$se <- sqrt(sum(x$weight^2 * x$var_o_minus_e))
     ans$z <- ans$estimation / ans$se
 
   } else if (inherits(weight, "mb")) {
     x <- x |> mb_weight(delay = weight$delay, w_max = weight$w_max)
 
     ans$parameter <- paste0("MB(delay = ", delay, ", max_weight = ", w_max, ")")
-    ans$estimate <- sum(res$o_minus_e * res$mb_weight)
-    ans$se <- sqrt(sum(res$var_o_minus_e * res$mb_weight^2))
+    ans$estimate <- sum(x$o_minus_e * x$mb_weight)
+    ans$se <- sqrt(sum(x$var_o_minus_e * x$mb_weight^2))
     ans$z <- ans$estimate / ans$se
 
   } else if (inherits(weight, "early_period")) {
     x <- x |> early_zero_weight(early_period = weight$early_period)
 
-    ans$parameter <- paste0("Xu 2017 with first ", early_period, " months of 0 weights")
-    ans$estimate <- sum(res$o_minus_e * res$weight)
-    ans$se <- sqrt(sum(res$var_o_minus_e * res$weight^2))
+    ans$parameter <- paste0("Xu 2017 with first ", weight$early_period, " months of 0 weights")
+    ans$estimate <- sum(x$o_minus_e * x$weight)
+    ans$se <- sqrt(sum(x$var_o_minus_e * x$weight^2))
     ans$z <- ans$estimate / ans$se
 
   }

R/wlr_weight.R

@@ -151,6 +151,6 @@ mb <- function(delay = 4, w_max = Inf) {
 #' ) |>
 #'   cut_data_by_event(125) |>
 #'   wlr(weight = early_zero(early_period = 2, fail_rate = fail_rate))
-early_zero <- function(early_period, fail_rate) {
+early_zero <- function(early_period, fail_rate = NULL) {
   structure(list(early_period = early_period, fail_rate = fail_rate), class = c("list", "early_period", "wlr"))
 }

tests/testthat/test-unvalidated-early_zero_weight.R

@@ -1,10 +1,11 @@
 test_that("early_zero_weight() with unstratified data", {
   # Example 1: Unstratified
   set.seed(123)
-  input <- sim_pw_surv(n = 200)
-  input <- cut_data_by_event(input, 125)
-  input <- counting_process(input, arm = "experimental")
-  output <- early_zero_weight(input, early_period = 2)
+
+  output <- sim_pw_surv(n = 200) |>
+    cut_data_by_event(125) |>
+    counting_process(arm = "experimental") |>
+    early_zero_weight(early_period = 2)
 
   observed <- output$weight
   expected <- rep(c(0, 1), c(15L, 110L))

vignettes/pvalue-maxcombo.Rmd → vignettes/maxcombo.Rmd

@@ -49,7 +49,9 @@ library(dplyr)
 library(gt)
 ```
 
-```{r}
+```{r, eval=FALSE}
+set.seed(123)
+
 x <- sim_fixed_n(
   n_sim = 1,
   timing_type = 5,
@@ -66,6 +68,8 @@ We begin with another simulation generated by `sim_pw_surv()`.
 Again, we use defaults for that routine.
 
 ```{r, message=FALSE, warning=FALSE, cache=FALSE}
+set.seed(123)
+
 s <- sim_pw_surv(n = 100)
 
 s |> 
@@ -133,15 +137,16 @@ x |> head() |> gt()
 Now we analyze the data with a MaxCombo with the logrank and FH(0, 1) and compute a $p$-value.
 
 ```{r, warning=FALSE, message=FALSE}
-x |>
-  maxcombo(rho = c(0, 0), gamma = c(0, 1))
+x |> maxcombo(rho = c(0, 0), gamma = c(0, 1))
 ```
 
 ## Simulation
 
 We now consider the example simulation from the `pvalue_maxcombo()` help file to demonstrate how to simulate power for the MaxCombo test. However, we increase the number of simulations to 100 in this case; a larger number should be used (e.g., 1000) for a better estimate of design properties. Here we will test at the $\alpha=0.001$ level.
 
 ```{r, cache=FALSE, warning=FALSE, message=FALSE, eval=FALSE}
+set.seed(123)
+
 # Only use cut events + min follow-up
 xx <- sim_fixed_n(
   n_sim = 100,
@@ -161,8 +166,10 @@ We note the use of `group_map` in the above produces a list of $p$-values for ea
 It would be nice to have something that worked more like `dplyr::summarize()` to avoid `unlist()` and to allow evaluating, say, multiple data cutoff methods.
 The latter can be done without having to re-run all simulations as follows, demonstrated with a smaller number of simulations.
 
-```{r, cache=FALSE, warning=FALSE, message=FALSE}
+```{r, cache=FALSE, warning=FALSE, message=FALSE, eval=FALSE}
 # Only use cuts for events and events + min follow-up
+set.seed(123)
+
 xx <- sim_fixed_n(
   n_sim = 100,
   timing_type = c(2, 5),