Sample size tools for various designs [Power / Sample Size]

posted by Helmut Homepage – Vienna, Austria, 2020-09-28 13:31 (854 d 19:27 ago) – Posting: # 21946
Views: 1,244

Hi Achievwin,

❝ share tools (R-, SAS code and excel spreadsheet) for computing Sample size from ISCV for Parallel, 2x2, 3x3 and 4x4 BE study designs) …

The CV and T/R-ratios are assumptions (or estimates if obtained from previous studies). Hence, sample size estimation (not computation or calculation), if you don’t mind. ;-)

I recommend the [image] package PowerTOST. For the implemented designs see here and there. You can also run scripts in the browser (see this post).
If you are dealing with a higher-order design, I recommend the “Two at a Time” approach instead of “All at Once” (pooled ANOVA). See the vignette. That means to estimate the sample size of the study like for a 2×2×2 design.

In a parallel design you get only the total (pooled) CV.
For the intra- (and inter-) subject CV you need a crossover.

If you want sumfink in M$ Excel, consider FARTSSIE which estimates the sample size based on the noncentral t-distribution. Note that the sample size for replicate designs is only approximate and for reference-scaling wrong (since no algebraic solution exist and therefore, simulations are required). For 2×2×2 studies you can also implement approximations based on the central t-distribution.1 However, I don’t recommend that because in borderline cases the sample size will be higher than necessary:

res <- data.frame(method = c("exact", "noncentral", "central"))
for (j in 1:nrow(res)) {
  res$n[j] <- sampleN.TOST(CV = 0.22, theta0 = 0.95, targetpower = 0.8,
                           method = res$method[j], details = FALSE,
                           print = FALSE)[["Sample size"]]
print(res, row.names = FALSE)
    method  n
     exact 22
noncentral 22

Some SAS code based on the noncentral t-distribution is given by Jones and Kenward.2 AFAIK, no code for reference-scaling is in the public domain. You are on your own – good luck and be prepared for extreme run­times.

❝ … also if we can compute ISCV or ANOVA CV from the confidence intervals.

Sure – if you know also the sample size and design. For the underlying algebra see this presentation (slides 26–30). Implemented in PowerTOST’s functions CVfromCI() / CI2CV(). See also the vignette. Example:

signif(CVfromCI(lower = 0.9800, upper = 1.1257, design = "2x2x4", n = c(62, 63)), 4)
# [1] 0.497

  1. Hauschke D, Steinijans VW, Diletti E, Burke M. Sample Size Determination for Bioequivalence Assessment Using a Multiplicative Model. J Pharmacokinet Biopharm. 1992; 20(5): 557–61. doi:10.1007/BF01061471.
  2. Jones B, Kenward MG. Design and Analysis of Cross-Over Trials. Boca Raton: Chapman & Hall, CRC Press; 3rd edition 2015.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

UA Flag
 Admin contact
22,477 posts in 4,708 threads, 1,603 registered users;
11 visitors (0 registered, 11 guests [including 5 identified bots]).
Forum time: 07:59 CET (Europe/Vienna)

The mediocre teacher tells.
The good teacher explains.
The superior teacher demonstrates.
The great teacher inspires.    William Arthur Ward

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz