prome (version 3.0.1.5)

blinding.cpe: Changing point estimator to correct bias due to unblinding in RCTs

Description

estimate the sham effect and correct effect size

Usage

blinding.cpe(x,group,guess)

Value

3 estimates, BI indices.

Arguments

x

outcome variable. numeric vector

group

group assignment. Coded as "0"=control and "1"=active/treatment. If 'group' is a factor, the first level will be treated as "control" arm. For example, if there are two values (ie. "ctrl" and "active"), "active" will be treated as the control arm. If the two levels are "control" and "treatment", "control" will be treated as the control arm.

guess

responses to the blinding survey question. The response corresponding to positive sham effect needs to be coded as "1", and the rest as "0". If the possible responses are "Active", "Control" and "I don't know", code "guess" as "1" for "Active" and "0" for the others if a subject responded "active" is expect to have positive sham effect. Otherwise, if a subject responded "active" is expect to have negative sham effect, code "guess" as "0" for "Active" and "1" for the others.

Details

To be added

Examples

# \donttest{
u1      = 5.5 # trt
u2      = 2.0 # ctrl
theta   = 3.2 # sham
sigma2  = 2.5   # v(rij)
ntreat  = 500      
nsham   = 500

beta0 = 1.0
beta1 = 2.0
beta2 = 1.0 # no contamination

Tind  = c(rep(1, ntreat), rep(0,nsham))  #treatment group indicator
u1v   = rep(u1,ntreat)
u2v   = rep(u2,nsham)
uv    = c(u1v,u2v)
tauv  = uv - rep(u2, ntreat+nsham)
r = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma2))
q = 1/(1 + exp(-(beta0 + beta1*Tind + beta2*(tauv+r))))
bernGen = function(qq){rbinom(1,1,qq)}
I = sapply(q,bernGen)
x = uv + theta*I + r   # fixed sham effect
## I have concerns about the error term(s). x.sham~N(theta,sigma.sham)?
sigma.sham = 1.5
r2 = rnorm(ntreat + nsham, mean = 0, sd = sqrt(sigma.sham))
x = (uv + r) + theta*I #+ r2   # fixed sham effect

out1 <- blinding.cpe(x=x,group=Tind,guess=I);
out1

##data(bd012)
##blinding.cpe(x=bd012$y, group=bd012$group,guess=bd012$guess)
##data(bd011)
##blinding.cpe(x=bd011$y, group=bd011$group,guess=bd011$guess)
##data(bd010)
##blinding.cpe(x=bd010$y, group=bd010$group,guess=bd010$guess)
# }