If it is real: results [RSABE / ABEL]
Hi Martin!
Oh yes, sure. Sorry for my stupidity.
I always thought he should work on his social skills.
Here the results of my sim’s for target α 0.05. Partial and full replicate designs, n=24/48.
The FDA’s method would lead to a large α-inflation (as presented by Detlew above) for CVWR ≤30%. This could be corrected by adjusting α. However, since scaling is only allowed for CVWR >30% nothing has to be done. Like the conventional TOST the test gets more conservative with increasing CV (and n).
The EMA’s method is another cup of tea. α-inflation is seen up to 45%! The test is closer to the nominal level than the FDA’s.
Interesting the behavior of the partial replicate with n=24.
Code (10–25 iterations to reach convergence, runtime on my machine 15–45 seconds depending on design, n, CV):
❝ IMHO: yes as this is a constant and should not dependent on the number of simulation runs.
Oh yes, sure. Sorry for my stupidity.
❝ Ps.: ... and yes its the same guy who is famous for publicly calling some approaches “bullshit”
I always thought he should work on his social skills.
Here the results of my sim’s for target α 0.05. Partial and full replicate designs, n=24/48.
The FDA’s method would lead to a large α-inflation (as presented by Detlew above) for CVWR ≤30%. This could be corrected by adjusting α. However, since scaling is only allowed for CVWR >30% nothing has to be done. Like the conventional TOST the test gets more conservative with increasing CV (and n).
The EMA’s method is another cup of tea. α-inflation is seen up to 45%! The test is closer to the nominal level than the FDA’s.
Interesting the behavior of the partial replicate with n=24.
Code (10–25 iterations to reach convergence, runtime on my machine 15–45 seconds depending on design, n, CV):
require(PowerTOST)
reg <- "EMA" # "EMA" for ABEL or "FDA" for scABE
CV <- 0.300 # intra-subject CV of reference
n <- 24 # total sample size
# in case of imbalanced sequences use a vector:
# e.g., c(n1,n2,n3)
des <- "2x3x3" # partial: "2x3x3" full: "2x2x4"
# for others see known.designs()
print <- TRUE # intermediate results
if (reg == "EMA") { # scABEL
ifelse (CV <= 0.5, GMR <- exp(0.7601283*CV2se(CV)),
GMR <- exp(0.7601283*CV2se(0.5)))
if (CV <= 0.3) GMR <- 1.25
} else { # RSABE
ifelse (CV > 0.3, GMR <- exp(0.8925742*CV2se(CV)),
GMR <- 1.25)
}
nsims <- 1e6
prec <- 1e-8 # precision of bisection algo
x <- 0.05 # target alpha
nom <- c(0.001, x) # from conservative to target alpha
lower <- min(nom); upper <- max(nom)
delta <- upper - lower # interval
ptm <- proc.time() # start timer
iter <- 0
while (abs(delta) > prec) { # until precision reached
iter <- iter + 1
x.new <- (lower+upper)/2 # bisection
if (reg == "EMA") {
pBE <- power.scABEL(alpha=x.new, theta0=GMR, CV=CV,
n=n, design=des, nsims=nsims)
} else {
pBE <- power.RSABE(alpha=x.new, theta0=GMR, CV=CV,
n=n, design=des, nsims=nsims)
}
if (print) { # show progress
if (iter == 1) cat(" i adj. alpha pBE\n")
cat(format(iter, digits=2, width=2),
format(x.new, digits=6, width=11, nsmall=7),
format(pBE, digits=6, width=9, nsmall=6), "\n")
if (.Platform$OS.type == "windows") flush.console()
}
if (abs(pBE - x) <= prec) break # precision reached
if (pBE > x) upper <- x.new # move upper limit downwards
if (pBE < x) lower <- x.new # move lower limit upwards
delta <- upper - lower # new interval
}
if (print) cat("run-time:", round((proc.time()[3]-ptm[3]), 1),
"seconds iterations:", iter, "\n")
cat("regulator:", reg, " CV:", CV, " n:", n, " design:", des,
"\ntarget alpha:", x, " adjusted alpha:", x.new,
" pBE:", pBE, "(empiric alpha)\n")
—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz
The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz
The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Complete thread:
- ‘alpha’ of scaled ABE? d_labes 2013-03-15 15:56 [RSABE / ABEL]
- ‘alpha’ of scaled ABE? Helmut 2013-03-15 17:27
- ’alpha’ of scaled ABE: design d_labes 2013-03-16 20:10
- N=24, 48 d_labes 2013-03-18 08:11
- adaptive design without adjustment martin 2013-03-25 20:47
- α adjustment? Helmut 2013-03-26 15:21
- iteratively adjusted alpha martin 2013-03-26 16:19
- throw away our sample size tables? Helmut 2013-03-26 18:02
- if real martin 2013-03-26 19:44
- regulators don’t care? Helmut 2013-03-26 21:18
- adjusting α Helmut 2013-08-11 16:19
- if real martin 2013-03-26 19:44
- target α in sim’s? Helmut 2013-03-27 15:38
- target α in sim’s? martin 2013-03-27 16:36
- If it is real: resultsHelmut 2013-03-28 23:20
- having alpha martin 2013-03-28 15:08
- fixed swr Helmut 2013-03-28 23:34
- fixed swr martin 2013-03-29 15:07
- Yes but no but yes but no but… Helmut 2013-03-29 16:10
- fixed swr martin 2013-03-29 15:07
- fixed swr Helmut 2013-03-28 23:34
- target α in sim’s? martin 2013-03-27 16:36
- throw away our sample size tables? Helmut 2013-03-26 18:02
- iteratively adjusted alpha martin 2013-03-26 16:19
- α adjustment? Helmut 2013-03-26 15:21
- adaptive design without adjustment martin 2013-03-25 20:47
- N=24, 48 d_labes 2013-03-18 08:11
- ’alpha’ of scaled ABE: design d_labes 2013-03-16 20:10
- ‘alpha’ of scaled ABE? Helmut 2013-03-15 17:27