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

Steve Chervitz stevechervitz@yahoo.com
Tue, 8 Jan 2002 01:35:23 -0800 (PST)


Hilmar Lapp wrote:

> Jason Stajich wrote:
> > > I'm implementing the
Bio::Search::HSP::GenericHSP and
> > Bio::Search::Hit::GenericHit objects, wanted to
clarify our use of e-value
> > and p-value.  I think we have been interchanging
them and considering p
> > and alias of evalue.  
> 
> Not really, and it shouldn't be. It's only that
BLAST at some
> point stopped reporting the p-value, and reported
the e-value
> instead.
> 
> The question is rather what significance() should
be. I named it
> to have somewhat an abstraction from what in a
particular hit
> and/or alignment tells you the significance. It
could even be a
> score sitting behind it. I.e., significance() is
(was meant to be)
> the general-purpose figure of significance, which an
individual
> implementation (inheriting class) can map to
whatever is
> appropriate

I have a similar method in BlastHit and BlastHSP
called signif() that
uses the P-value if it is defined, otherwise it uses
E-value. My aim
was to allow someone to write code that will work with
a variety of
Blast report formats (P and E are equivalent below
0.01 or
so, where most people really operate).

P-values are still used in WU-BLAST (at least in the
free version 05-Feb-1998).

Here's the POD from my BlastHit::signif():

  : The signif() method provides a way to deal with
the fact that
  : Blast1 and Blast2 formats (and WU- vs. NCBI-BLAST)
differ in 
  : what is reported in the description lines of each
hit in the 
  : Blast report. The signif() method frees any client
code from 
  : having to know if this is a P-value or an Expect
value, 
  : making it easier to write code that can process
both 
  : Blast1 and Blast2 reports. This is not necessarily
a good thing, 
  : since one should always know when one is working
with P-values or
  : Expect values (hence the deprecated status).
  : Use of expect() is recommended since all hits will
have an Expect value.

Note that it's not officially deprecated yet, since I
don't generate a
warning.

Steve



__________________________________________________
Do You Yahoo!?
Send FREE video emails in Yahoo! Mail!
http://promo.yahoo.com/videomail/