Bioequivalence and Bioavailability Forum

Dear Detlew,

❝ ❝ Maybe it’s safer to replace […]

❝ This is only necessary for a function definition outside package. Within the next release of PowerTOST, which will introduce the sensitivity analysis, the correct version of power2.TOST() is of course available.

Yes, sure.

❝ ❝ Most input errors are already trapped by PowerTOST. […] otherwise uniroot() throws an error.

❝ Done. Do you know the real reason behind that error?

0.05 works and 0.049 doesn’t. Especially looking at the output-vector and the plots not the slightest idea why.

❝ ❝ Consider adding ...]

❝ But ran into trouble doing so with some inexplicable (for me) errors. Seems have to study how to work with the argument '...'.

I’ve tried it myself before. I was hoping that you know better than I do. Not important at the moment.

❝ ❝ Sneak Preview

❝

❝ Funny .

❝ Especially the deviation for the CV!

❝ Seems we don't have to care for any deviation here. In the first look surprising.

That’s really nice to show the idea behind scaling. The CV is “dealt with” and the other parameters behave like in TOST.

❝ Some adaption of the argument names seems necessary to me to be in accordance with the naming within PowerTOST […]

Sure.

❝ I would suggest the function name to be something like psa.ABE() (power sensitivity analysis for ABE) or psa.TOST() to account for the upcoming things seen in your Sneak Preview.

I would suggest to avoid the term “sensitivity analysis”. Looking at the plots novices might get the false impression that effects may occur simultaneously “CV can go up to x%, the GMR to y, and with y drop-outs I have still 80% power”. We should make that very clear in the man-page. My suggestion is pa.TOST() (power analysis for TOST). In my code for scaling I have included both RSABE and ABEL, therefore, pascABEl() and paRSABE() – which would be inline with PowerTOST’s naming conventions is not an option. It should be easy to include FDA’s NTID-scaling as well. Maybe something like paScaledBE()?