Bioequivalence and Bioavailability Forum

Dear Martin!

I must controvert Helmut's modification of Wittgenstein. There are methods out there in the statistical universe by which we can "rechnen" with data having values known only to the extent of being lower than a certain value (or higher than a certain value or known only in an interval).
I would suggest you to look for statistical methods for censored / truncated data.

Unfortunately I have no personal experience / knowledge with such methods. Thus I can't direct you to a ready to use solution.

But here is a point of view with literature from the field of environment sciences.

D. Helsel
Much Ado About Next to Nothing:
Incorporating Nondetects in Science
The Annals of Occupational Hygiene

Dennis Helsel has also written a book:
"Nondetects And Data Analysis: Statistics for Censored Environmental Data"
Wiley 2004
ISBN: 978-0-471-67173-2

The methods described are implemented in the R package NADA. See http://www.practicalstats.com/nada/downloads_files/NADAforR_Examples.pdf for examples.

ing for "non-detects statistics censored data" will give you much more interesting material. Among them this one showing that substitution of LOQ is not the best idea.

Hope this helps.