Phase III press release that I read at
biz.yahoo.com
seems to be missing some crucial data: while they report
p-values of 0.075 and 0.033
for two types of statistical analysis of the data
which are (depending on your criterion, since some require 0.05)
significant for the 500mg tablet formulation, they don't report
the magnitude of the effect, which is important in judging these
results. Also, as the they point out "Due to the four arms in the study, the statistical significance of these values
must be discussed with the FDA." This is probably because, when you
study several different formulations, and find significant effect with
one, but not with several others, it is, overall, more likely that the
significant effect of the one formulation was due to chance, for the
same reason that if you repeat a study
over and over, you'll eventually get a "significant" result once
by chance even if there is no real effect.
Not that this is necessarily the case here. It would help to know
the size of the effect, both in judging both the importance of
the significant result (which could just support the existence of a
tiny, but real, effect) and in guesstimating whether these p-values
are associated with a real effect, or statistical fluctuations.
(I'd rather see a large effect at even slightly lower p-value, for
all formulations, than a small effect in all formulations,
"statistically significant" in some.)
Suspect that's why market reaction has been underwhelming. I don't own this stock, but I'm interested because Mike Burke does.
What's "Taipan newsletter", a tacky internet stock-pumping operation?
Cheers,
HB |
