Re: [ccp4bb]: sigmaa and HL coefficients

[Randy J Read <rjr27@cam.ac.uk>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

On Tue, 7 Aug 2001 10:34:31 +0100, Kevin Cowtan <cowtan@ysbl.york.ac.uk> wrote:
>
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> Actually, this sigmaa bug finally prompted me to look at SFTOOLS, and I wish 
> I had done so much sooner. It will convert phase+fom to HL coeffs (HLCONV), 
> and add HL coeffs (CALC).
>
> So you can do everything MIR2 does in SFTOOLS without patching your existing 
> code.
>

As much as I think everyone should be familiar with all the wonderful
things you can do in SFTOOLS, I have to disagree with part of what Kevin
seems to be suggesting here.  If you do indeed have two sets of HL coeffs,
you can use SFTOOLS to add them up with four CALC commands.  But please
don't use the HLCONV option to convert MIR phase and fom to A and B
coefficients, except as a last resort when you've lost the original data!
As SFTOOLS says when you give the command HLCONV help, you should only use
HLCONV when your phase information is unimodal (i.e. phases from a
molecular model or density modification).  When it is bimodal, you lose a
lot of information by representing it as a unimodal distribution.  As an
extreme case, if you have SIR phases, there are two peaks of equal height,
which may have very low probability between them.  The HLCONV option will
give you a single peak halfway between the two peaks from the SIR
distribution, with high probability where the original one was low, and low
probability where the original one had two high peaks.

Randy Read