Following Ian's excellent comment, do I understand correctly that
2mFo-DFc is the maximum likelihood estimate of the full model map (i.e.
the best map possible) or it's simply modification of 2Fo-Fc map where
plain Fo/Fc are replaced by their maximum likelihood estimates?  Other
words, is k=2 the maximum likelihood estimate of the best approximation
of the true map in the following form

DFc + k*(mFo-DFc)

Ed.

On Wed, 2010-09-01 at 10:49 +0100, Ian Tickle wrote:
> On Wed, Sep 1, 2010 at 4:26 AM, Ed Pozharski <[email protected]> wrote:
> > The
> > reason you see the missing region in (2mFo-DFc) map is because it is
> > effectively the sum of model map (mFo) which shows you the parts of the
> > model you have already placed and difference map (mFo-DFc) which shows
> > you the regions which are still missing.
> 
> This is not true.  The 'model map' (i.e. the map calculated from the
> model) is obviously the one with coefficient DFc.  The mFo map
> represents the model (i.e. the structure already placed) + *half* of
> the missing structure (represented by mFo-DFc), for acentric
> reflections.  To get the 'minimally biased' map you have to make it up
> by adding the other half of the missing structure so we have (for
> acentrics):
> 
> 2mFo-DFc = DFc + (mFo-DFc) + (mFo-DFc)
>                 = DFc + 2(mFo-DFc)
> 
> For centrics mFo represents the model + *all* of the missing
> structure, so in that case no further contribution is needed,
> 
> We had this discussion a while back: it seems to me that it is
> precisely this confusion that is engendered by thinking in terms of
> 2mFo-DFc = mFo + (mFo-DFc).
> 
> Cheers
> 
> -- Ian

-- 
"I'd jump in myself, if I weren't so good at whistling."
                               Julian, King of Lemurs

Reply via email to