trial: Run a simulated trial

In maj-biostat/coadministr: Trial simulator for coadministration design

Description

Runs the all analyses for a coadministration trial

Usage

trial <- function(
    scenario = 0,
    looks = c(200, 300, 400, 500), n_per_yr = 2000, delta_ni = 0.2, 
    beta = NULL, sd_id = 0.5, p_miss = 0.0,
    Ncoad = 2, Nbrand = 2, Ncohort = 2, 
    a0 = 1, b0 = 1, eta0 = 0.01, keep_post_sampls = FALSE
    )

Arguments

scenario	an id for designating the current scenario
looks	sample size to run interim and final analyses
n_per_yr	enrollment per year
delta_ni	non-inferiority margin on probability scale
beta	simulation parameters, see pr_adverse_evt
sd_id	subject level variation
p_miss	probability of missingness
Ncoad	number of treatments - presently fixed to 2
Nbrand	number of brands - presently fixed to 2
Ncohort	number of cohorts - presently fixed to 2
a0	hyperparameter for beta prior
b0	hyperparameter for beta prior
eta0	hyperparameter for exponential prior
keep_post_sampls	returns posterior samples (or not)

Details

After setting up some preliminary data structures, this function simulates the data for this trial then proceeds through the interim analyses. At each interim, the data accrued to date is used to model the adverse events for the coadministered covid and flu and the coadministered covid and placebo for each brand and cohort (where cohort denotes being enrolled at either dose 1 of the covid vaccine or dose 2 of the covid vaccine).

After all the interim analyses have been completed, a final analysis is run on all the data after followup is completed then the trial dataset, analysis summaries and config are returned to the user.

Best used in parallel per the example.

Value

List of data and parameters used in the trial and trial analysis results

d	Simulated dataset used for trial
lpar	Simulation parameters
lr	Summaries of analysis results
lpost	Posterior samples (if requested) otherwise NULL

Author(s)

Mark

Examples

## Not run: 
library(data.table)
library(coadministr)
# Same incr in Pr AE in CVD+FLU regardless of whether cvd dose 1 and dose 2
# cvd+flu combination increases pr AE by 0.1 for AZ
# cvd+flu combination increases pr AE by 0.15 for PF
b <- pr_adverse_evt(pae_pbo = 0.02, pae_flu = 0.05, 
                    pae_coad_az = 0.1, pae_coad_pf = 0.15,
                    pae_az1 = 0.5, pae_az2 = 0.4,
                    pae_pf1 = 0.3, pae_pf2 = 0.5)
d <- get_data(J = 1000,
              beta = b,
              sd_id = 0.5,
              n_per_yr = 1000,
              p_miss = 0.1)
results <- coadministr::trial(
   scenario = 1,
   looks = seq(500, 1000, by = 100),
   n_per_yr = 2000,
   delta_ni = 0.2, 
   beta = b,
   sd_id = 0.5,
   p_miss = 0.1,
   Ncoad = 2,
   Nbrand = 2,
   Ncohort = 2,
   a0 = 1,
   b0 = 1,
   eta0 = 0.01,
   keep_post_sampls = TRUE
 )
 # Note that results$lpost[[1]]$pflu[, , 1] contains draws for AZ coad CVD+FLU 
 # at covid dose 1 and dose 2 in cols 1 and 2 respectively.
 # ditto for results$lpost[[1]]$pflu[, , 2] but for PF
 # ditto for results$lpost[[1]]$ppbo[, , 1] but for AZ CVD+PBO
 # ditto for results$lpost[[1]]$ppbo[, , 2] but for PF CVD+PBO
 str(results$lpost[[1]])
 post_to_data_table(results$lpost[[1]])

## End(Not run)
## Not run: 
library(parallel)
library(data.table)
ncores <- 10
results <- parallel::mclapply(X=1:10, mc.cores = ncores, FUN=function(x) {
 # should use tryCatch block
 b <- pr_adverse_evt(pae_pbo = 0.02, pae_flu = 0.05, pae_coad = 0.1)
 coadministr::trial(
   scenario = 1,
   looks = seq(500, 1000, by = 100),
   n_per_yr = 2000,
   delta_ni = 0.2, 
   beta = b,
   sd_id = 0.5,
   p_miss = 0.1,
   Ncoad = 2,
   Nbrand = 2,
   Ncohort = 2,
   a0 = 1,
   b0 = 1,
   eta0 = 0.01
 )
})

## End(Not run)

maj-biostat/coadministr documentation built on Dec. 21, 2021, 1:44 p.m.