From 93e5e269798ace159a481adba1d1f0ced3a81233 Mon Sep 17 00:00:00 2001
From: Chris Jackson <chris.jackson@mrc-bsu.cam.ac.uk>
Date: Wed, 8 Mar 2023 12:12:48 +0000
Subject: [PATCH] Bugfix for EVSI output with multiple n and WTP

---
 R/evsi.R                   | 16 +++++++++++-----
 tests/testthat/test_evsi.R | 28 +++++++++++++++++++++++-----
 2 files changed, 34 insertions(+), 10 deletions(-)

diff --git a/R/evsi.R b/R/evsi.R
index fb31fd6..4d959ab 100644
--- a/R/evsi.R
+++ b/R/evsi.R
@@ -22,18 +22,24 @@
 ##'   parameters: the probability of the outcome in arm 1 and 2 respectively.
 ##'   The sample size is the same in each arm, specifed in the \code{n} argument
 ##'   to \code{evsi()}, and the binomial outcomes are named \code{X1} and
-##'   \code{X2} respectively.
+##'   \code{X2} respectively.  
 ##'
 ##'   \code{"normal_known"} A study of a normally-distributed outcome, with a
 ##'   known standard deviation, on one sample of individuals.  Likewise the
 ##'   sample size is specified in the \code{n} argument to \code{evsi()}.  The
 ##'   standard deviation defaults to 1, and can be changed by specifying
 ##'   \code{sd} as a component of the \code{aux_pars} argument, e.g.
-##'   \code{evsi(..., aux_pars=list(sd=2))}.
+##'   \code{evsi(..., aux_pars=list(sd=2))}.   
 ##'
 ##'   Either \code{study} or \code{datagen_fn} should be supplied to
 ##'   \code{evsi()}.
 ##'
+##'   For the EVSI calculation methods where explicit Bayesian analyses of the
+##'  simulated data are performed, the prior parameters for these built-in studies
+##'  are supplied in the \code{analysis_args} argument to \code{evsi()}.  These
+##'   assume Beta priors for probabilities, and Normal priors for the mean of a
+##' normal outcome. . 
+##'
 ##'
 ##' @param datagen_fn If the proposed study is not one of the built-in types
 ##'   supported, it can be specified in this argument as an R function to sample
@@ -186,8 +192,8 @@
 ##'   for the first arm,  `a2` and `b2`: Beta shape parameters for the prior for
 ##'   the second arm.
 ##'
-##'   `study="normal_known"`: `prior_mean`, `prior_sd` (prior mean and standard
-##'   deviation) and `sampling_sd` (SD of an individual-level normal
+##'   `study="normal_known"`: `prior_mean`, `prior_sd` (mean and standard deviation 
+##'   deviation of the Normal prior) and `sampling_sd` (SD of an individual-level normal
 ##'   observation, so that the sampling SD of the mean outcome over the study is
 ##'   `sampling_sd/sqrt(n)`.
 ##'
@@ -312,7 +318,7 @@ evsi_npreg <- function(outputs, inputs, datagen_fn, pars, n, method=NULL, se=FAL
         rownames(res[[i]]) <- NULL
     }
     resall <- do.call("rbind", res)
-    resall <- cbind(n=n, resall)
+    resall <- cbind(n=rep(n,each=nrow(res[[1]])), resall)
     if (check){
       attr(resall, "models") <- lapply(res, function(x)attr(x, "models"))
       names(attr(resall,"models")) <- as.character(resall$n)
diff --git a/tests/testthat/test_evsi.R b/tests/testthat/test_evsi.R
index ca870b8..9e05619 100644
--- a/tests/testthat/test_evsi.R
+++ b/tests/testthat/test_evsi.R
@@ -77,11 +77,19 @@ test_that("EVSI with built-in study designs", {
 
 test_that("EVSI with built-in study designs: IS method", {
     set.seed(1)
-    expect_equal(
-        evsi(chemo_nb, chemo_pars, study="trial_binary", pars=c(pi1, pi2), 
-             method="is", nsim=1000, verbose=FALSE)$evsi
-      , 
-        789.1255, tol=0.01)
+  expect_equal(
+    evsi(chemo_nb, chemo_pars, study="trial_binary", pars=c(pi1, pi2), 
+         method="is", nsim=1000, verbose=FALSE)$evsi
+    , 
+    789.1255, tol=0.01)
+  
+  ## can we include extra cols 
+  chemo_pars$xextra <- 1 
+  expect_equal(
+    evsi(chemo_nb, chemo_pars, study="trial_binary", pars=c(pi1, pi2), 
+         method="is", nsim=1000, verbose=FALSE)$evsi
+    , 
+    789.1255, tol=0.01)
 })
 
 test_that("EVSI with multiple sample sizes", { 
@@ -104,3 +112,13 @@ test_that("EVSI with multiple sample sizes", {
              verbose=FALSE, method="is", nsim=1000) 
   expect_equal(e2$evsi[1], 571.2093, tol=1e-04)
 })
+
+test_that("EVSI with multiple sample sizes and CEA output", { 
+  set.seed(1)
+  e1 <- evsi(chemo_cea, chemo_pars, study="binary", n=100, pars=c(pi2), verbose=FALSE)
+  e2 <- evsi(chemo_cea, chemo_pars, study="binary", n=1000, pars=c(pi2), verbose=FALSE) 
+  set.seed(1)
+  e12 <- evsi(chemo_cea, chemo_pars, study="binary", n=c(100,1000), pars=c(pi2), verbose=FALSE) 
+  expect_equal(e12$evsi, c(e1$evsi,e2$evsi), tol=1e-03)
+  expect_equal(e12$n, c(e1$n,e2$n))
+})