trying to get the cgd vignette to run but it crashes. TODO

NAMESPACE

@@ -1,19 +1,15 @@
 # Generated by roxygen2: do not edit by hand
 
-export(bmcm_stan)
-export(cf_forest_cutpoint)
-export(cf_forest_plot)
-export(create_stancode)
-export(prep_latent_data)
+import(adegenet)
 import(dplyr)
+import(epicontacts)
 import(ggplot2)
 import(grid)
 import(gridExtra)
 import(purrr)
 import(rstanarm)
 import(survival)
 import(tidybayes)
-importFrom("epicontacts,",adegenet)
 importFrom(dplyr,mutate)
 importFrom(glue,glue)
 importFrom(glue,glue_collapse)

R/bmcm_stan.R

@@ -42,7 +42,6 @@ bmcm_stan <- function(input_data,
                       t_max = 70,
                       ...) {
   call <- match.call()
-
   rtn_wd <- getwd()
   setwd(here::here("inst/stan"))
   on.exit(setwd(rtn_wd))
@@ -70,7 +69,6 @@ bmcm_stan <- function(input_data,
     }}
 
   formula_latent <- parse_formula(formula, input_data, family = distns)
-
   if (is_pooled(formula_cure)) {
     formula_cure$cf_idx <- 1L
     formula_cure$cf_name <- "pooled"

R/cf_forest_cutpoint.R

@@ -7,8 +7,7 @@
 #' @param save_name File name appended text
 #' @param is_hier Is hierarchical model (or separate)? i.e. whether to use a global parameter
 #'
-#' @importFrom epicontacts, adegenet
-#' @import dplyr ggplot2 tidybayes purrr
+#' @import dplyr ggplot2 tidybayes purrr epicontacts adegenet
 #'
 #' @return
 #' @export

R/default_prior_cure.R

@@ -13,7 +13,6 @@
 default_prior_cure <- function(formula_cure,
                                prior_cure = list(),
                                bg_model = 2) {
-
   nTx <- formula_cure$fe_nlevels[1]
   n_groups <- formula_cure$n_groups
   nvars <- formula_cure$nvars

R/parse_formula.R

@@ -8,7 +8,6 @@ parse_formula <- function(formula, data, family = NA) {
   # if (!inherits(formula, "formula")) {
   #   stop("'formula' must be a formula.")
   # }
-
   formula <- as.formula(formula)
 
   # all variables of entire formula
@@ -32,15 +31,18 @@ parse_formula <- function(formula, data, family = NA) {
   bars      <- lme4::findbars(formula)[[1]]
   re_parts  <- split_at_bars(bars)
 
-  mf <- model.frame(lme4::subbars(formula), data = data)
+  # Substitute the '+' function for the '|' function
+  formsub <- lme4::subbars(formula)
+  mf <- model.frame(formsub, data = data)
 
   # names of variable without event type
   fe_vars <- attr(terms(fe_form), "term.labels")
   fe_nvars <- length(fe_vars)
 
   re_nlevels <- length(unique(mf[, re_parts$re_group_var]))
 
-  mf[, re_parts$re_group_var] <- as.factor(mf[, re_parts$re_group_var])
+  if (!is.null(re_parts$re_group_var))
+    mf[, re_parts$re_group_var] <- as.factor(mf[, re_parts$re_group_var])
 
   fe_nlevels <- apply(mf[, fe_vars, drop=FALSE], 2, \(x) nlevels(as.factor(x)))
 

vignettes/cgd_analysis.Rmd

@@ -67,7 +67,7 @@ Drop any levels which only appeared in the data we've just removed and convert t
 
 ```{r filter-center}
 cgd_center <- cgd_first |>
-  # filter(center %in% c("Amsterdam", "NIH", "Univ. of Zurich", "Scripps Institute")) |>
+  filter(center %in% c("Amsterdam", "NIH", "Univ. of Zurich", "Scripps Institute")) |>
   droplevels() |> 
   mutate(center_id = as.numeric(center),
          hos.cat_id = as.numeric(hos.cat),
@@ -126,12 +126,13 @@ out <-
     formula = "Surv(time=time, event=status) ~ 1",
     cureformula = "~ treat + (1 | center_id)",
     family_latent = "exponential",
-    prior_cure = NA,
+    # prior_cure = NA,
     centre_coefs = TRUE,
     bg_model = "bg_fixed",
     bg_varname = "rate",
     bg_hr = 1,
-    t_max = 365)
+    t_max = 400)
+    # t_max = 365)
 ```
 
 ### By hospital category

vignettes/how_to_use.Rmd

@@ -21,7 +21,7 @@ This is a basic introduction to how to use `rstanbmcm` to fit Bayesian mixture c
 
 ## Data
 
-We will use the Checkmate 067 study data set.
+We will use the Checkmate 067 study data set @ref.
 The data have already been arranged in to the correct format and saved within the package so we can load it as follows.
 
 ```{r eval=FALSE}
@@ -73,34 +73,20 @@ options(mc.cores = parallel::detectCores() - 1)
 ```
 
 Now we are ready to do the model fitting.
-There are 3 options to use.
-
-* `bmcm_joint_stan_string`: Constructs the Stan code on-the-fly, meaning it is flexible and can run all distributions combinations. It calls the Stan file directly from R without pre-compiling. This is useful for development.
-* `bmcm_joint_stan_file()`: Same as above but for pre-written code. Only available for same-distribution OS-PFS pairs.
-* `bmcm_joint_stan()`: Uses the pre-compiled Stan code. Quicker to run. Like `bmcm_joint_stan_file()` only for a subset of OS-PFS distribution pairs.
-
-An example call to `bmcm_joint_stan_string` is given below.
-We recommend using this because it is the most widely-applicable.
-The other functions can be used for testing.
 
 ```{r eval=FALSE}
 out <-
-  bmcm_joint_stan_string(
-    input_data = surv_input_data,
-    model_os = i,
-    model_pfs = k,
-    tx_name = j,
-    params_cf =
-      list(mu_cf_gl = array(-0.8, 1),
-           sigma_cf_gl = array(2, 1),
-           sd_cf_os = array(0.5, 1),
-           sd_cf_pfs = array(0.5, 1))
-    cf_model = 3,
-    joint_model = FALSE,
-    bg_model = 2,
-    warmup = 100,
-    iter = 1000,
-    thin = 10)
+  bmcm_stan(
+    input_data = input_data,
+    formula = "Surv(time=time, event=status) ~ 1",
+    cureformula = "~ treat + (1 | center_id)",
+    family_latent = "exponential",
+    # prior_cure = NA,
+    centre_coefs = TRUE,
+    bg_model = "bg_fixed",
+    bg_varname = "rate",
+    bg_hr = 1,
+    t_max = 400)
 ```
 
 ### Explanation of function arguments