findDTL: Find Multi-Outcome Two-Stage Drop-the-Loser designs

In martinlaw/moms: Creates Multi-Outcome Multi-Stage Trials With General Number of Efficacious Outcomes

Description

This function finds multi-outcome, two-stage drop-the-loser designs that declare trial success when a specified number of outcomes show promise. This function uses simulation.

Usage

findDTL(
  K,
  Kmax,
  m,
  alpha.k,
  power,
  corr.mat = NULL,
  vars = NULL,
  corr.scalar = NULL,
  delta0,
  delta1,
  delta.true = NULL,
  cp.l,
  cp.u,
  n.min,
  n.max,
  working.outs = NULL,
  nsims = 1000
)

Arguments

K	Number of outcomes
Kmax	Maximum number of outcomes permitted in stage 2
m	Number of outcomes required to show promise for trial success
alpha.k	The desired type-I error-rate.
power	The desired power.
corr.mat	A square matrix of the correlations between outcomes. Must be K-dimensional and have 1's on the diagonal.
vars	A vector of outcome variances. If single value is entered, it is used for all outcomes with a warning.
corr.scalar	A scalar of the correlation between outcomes. If entered, it is used for all correlations with a warning.
delta0	A vector of anticipated lower effect sizes. If a single value is entered, it is used for all outcomes with a warning.
delta1	A vector of anticipated upper effect sizes. If a single value is entered, it is used for all outcomes with a warning.
delta.true	Optional. A matrix of true effect sizes (with number of columns==K). If only 2 columns are supplied, will take delta.true1 as true delta for all working outcomes and delta.true2 as true delta for all non-working outcomes.
cp.l	The lower bound for conditional power.
cp.u	The upper bound for conditional power.
n.min	The minimum sample size to search over.
n.max	The maximum sample size to search over.
working.outs	A vector of the indices of outcomes that are taken to be the "working" or "best-performing" outcomes for the purposes of calculating the sample size. If not given, the first m outcomes will be used, with a warning.
nsims	The number of trials simulated. Default is 1000.

Details

if delta.true is used, an additional list element is returned, true.results, containing the operating characteristics of the obtained design taking into account the true effect sizes supplied.

Value

The function returns a list of length two The first element, input, contains the values inputted into the call. The second element, results, gives the final and interim stopping boundaries and the operating characteristics.

Examples

findDTL(K=4, Kmax=3, m=2, vars=c(1, 1.01, 2, 1.5), delta0=0.1, delta1=0.4, alpha.k=0.05, cp.l=0.3, cp.u=0.95, n.min=10, n.max=40, power=0.8, corr.scalar=0.4, working.outs=c(1,2))

m1 <- matrix(NA, 4, 4)
m1[lower.tri(m1, diag=F)] <- vec
m1 <- t(m1)
m1[lower.tri(m1, diag=F)] <- vec
diag(m1) <- 1
findDTL(nsims = 1e3, K=4, Kmax=3, m=2, vars = c(1, 1.01, 2, 1.5), delta0 = 0.1, delta1 = 0.4, alpha.k = 0.05, cp.l = 0.3, cp.u = 0.95, n.min = 10, n.max = 40, power = 0.8, corr.mat = m1, working.outs=c(1,2))

martinlaw/moms documentation built on June 18, 2021, 11:07 a.m.