Bioequivalence and Bioavailability Forum

Hi Hötzi,

❝ Say, you dosed N subjects and are interested in post hoc power of a fraction f of this sample size (e.g., of eligible subjects).

Accusing me of being a member of the post-hoc power fan club? I think my membership was discontinued around 2007.

For me, pre-study power considerations led me to look into this. Like checking a range of prospective attrition rates to evaluate the actual power.
Before, not after.

I have to say the bug is very minor. I have no relevant real-life (non-simulation) cases where this would make an much meaningful difference - extreme cases with unfavourable combinations of small sample sizes and high variabilities perhaps but most of what I do is not that extreme. A difference of e.g. 0.2% or whatever in a power estimation is to me not of particular relevance since in BE there are various assumption which can be violated and which may have unknown consequences for the magnitude of power.
Dasatinib comes to mind.

Add to that, that an unfortunate habit in the industry is to adjust assumptions to make the sample size / power meet a target, not the other way around. So, I would not invest too much too much time on fixing this one. PowerTOST is a great tool and the matter here does not in any way detract from its value. You and Detlew and possibly other people involved are doing a great job. Muchas gracias.