Outlier Tests [Outliers]
Thankyou for the reply.
❝ and thank for your patience
❝ I don't think this is an area covered very well by guidelines and I am not entirely sure there is a need either.
Yes... there is no clear guidelines on this topic.
❝ If you are really hardcore you can look at distribution theory for the points on the time series and then derive critical values; this is in line with the thinking behind studentized residuals for normally distributed data. I have no idea if this would be accepted, but it would at least be very stringent.
❝ If you are softcore, take one step back and ask your self what the impact of outliers is: They increase variability. And for that you have guidance or at least some regulatory practice and precedent to lean on.
❝ When variability goes up, the classical f2 comparison may not be valid and you'll need to provide proof via bootstrapping of f2 that the profiles are similar. Mahalanobis distance was also in play until recently but it went out of fashion. It was simply too difficult to spell and pronounce for Europeans, I believe .
Our in vitro studies are not dissolution studies. These are in vitro permeation studies. In these studies variability is high when compared to in vivo studies. Sample size is small when compared with in vivo studies. In this case, every value will impact the study results. So how can we deal this data, when outliers are present in the data?
One more thing, generally FDA guidance allows us to prove BE on log transformed data. So, on which data outlier test will be performed(either original scale or log scale)?
Thoughts on this topic is highly appreciated.
Thankyou in advance.