Re[2]: [S] "Sums of squares", the taxonomy issue again.

Chris Barker 415-852-3152 (
Wed, 29 Apr 1998 12:28:05 -0700 (PDT)

On re-reading the original note I wasn't sure what exactly was meant by saying
" > But we had to
> implement them in our [lm+aov] procedure, because
> U.S. government agencies mandate use of them."
Perhaps more clarification of the original discussion would help to explain
specific examples of mandates.

That said, I can easily imagine that various government agencies would like to
have certain methods of calculation available, just like anyone else.

I am in the pharmaceutical industry, and bioequivalence aside I am not aware of
anyplace where the FDA has written a guideline or procedure stipulating that a
certain type of sums of squares be computed for clinical trials in drug
development. It is my understanding that the creation of the guidelines for
bioequivalence did involve both government and industry statisticians.

However, for most of the clinical trials I have worked on the
"understanding" or folklore has
been that the preferred sums of squares is type III. The best advice is that
you tell FDA way in advance
what sort of sums of squares you are proposing to use for an application. They
in turn are free to analyze the data in whatever way they think is appropriate.
I've also found that industry frequently consults with statistical experts in
order to "get things right".

Chris Barker
Principal Pharmacoeconomic Statistician
Roche Pharma business

This message was distributed by To unsubscribe
send e-mail to with the BODY of the
message: unsubscribe s-news