[Bioperl-l] p-value, e-value

Aaron J Mackey Aaron J. Mackey" <amackey@virginia.edu
Tue, 8 Jan 2002 10:41:19 -0500 (EST)


On Tue, 8 Jan 2002, Steve Chervitz wrote:

> (P and E are equivalent below 0.01 or so, where most people really
> operate).

Just to help clarify (or confuse) the issue a bit more: often the P value
is the "per pairwise alignment probability", and the E value is the P
value times the size of the database (i.e. the "expectation of getting
this P, given that you did N pairwise alignments").  This E can then of
course be converted back into a P = 1- exp(-E), which, as Steve says, is
not very different when E < 0.01.  But the difference between the pairwise
P and the database E is quite large and important across all ranges of P.

-Aaron