Randomization for Replicate design studies [Design Issues]
❝ Once of the clients want to conduct the partial replicate BE study with two test products (T1 & T2) in comparison with reference product (R).
More information, please. Will this be a pilot study where you want to select one of the candidates or a pivotal study (say, T1 is a tablet, T2 a capsule and you want to market both)?
- For the former see this presentation (slide 22).
- For the latter employ the same design but adjust for multiplicity (i.e., 95% CI, slide 23).
If you expect a CVwR <43% the Type I Error might be inflated. You can iteratively adjust α (see here and there) with the R-package
scABEL.ad(alpha = 0.025, design = "2x3x3", ...), where in
...you provide the observed CVwR and sample size (in the case of drop-outs: subjects/sequence). In case of unbalanced sequences use the argument
sdsims = TRUE(slower but more accurate).
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
The quality of responses received is directly proportional to the quality of the question asked. 🚮