Hi Jay,

there have been a few discussions about sigma/rms levels, absolute
values (e/A*3), significance levels etc on ccp4bb. See e.g.

  [log in to unmask]" target="_blank">http:[log in to unmask]
  [log in to unmask]" target="_blank">http:[log in to unmask]

which are the ones I could find again (I'm sure there are others).

One example: imagine you would know the true value of a distribution,
say on 17 points. You then start to model it (this corresponds to your
initial stages of refinement). After some cycles, you end up with a
model you're happy with.

It could look like this:

 point  true   start     end
 -----------------------------
   1    1.0     1.1     1.00
   2    1.1     1.2     1.10
   3    1.5     1.4     1.52
   4    2.0     2.3     2.01
   5    2.7     2.9     2.72
   6    3.0     3.5     3.02
   7    4.0     4.2     4.04
   8    6.0     6.3     6.04
   9    9.0    10.0     9.09
  10    6.0     6.3     6.04
  11    4.0     4.2     4.04
  12    3.0     3.5     3.02
  13    2.7     2.9     2.73
  14    2.0     2.3     2.01
  15    1.5     1.4     1.52
  16    1.1     1.2     1.10
  17    1.0     1.1     1.00

Now plot those distributions (see attached file) and calculate
mean/sigma of the differences (between the current model and the true
values):

  start-true : <mean> =    0.2471  with  sd =    0.2476
    end-true : <mean> =    0.0235  with  sd =    0.0219

You will see that point 9 always corresponds to a '3 sigma' difference
peak - even if your model at the end is pretty much perfect (R-factor
at the beginning is 9% and at the end down to 1%). Just because a
distribution will always give you a mean and a sigma it doesn't mean
that you need to chase after those '3 sigma' peaks forever.

At least that is how I visualise the sigma/rms/electrons issue for me
...

Cheers

Clemens

On Wed, May 19, 2010 at 12:13:54AM +0100, Jay Pan wrote:
> Hello Everyone,
> 
> I have a reasonably well fitted electron density map through molecular replacement. However, there is always some red region left no matter how hard I tried when the mtz file is loaded into Coot. Is this because my model is still not good enough or it???s natural to most model fittings. In another word, should I be worried about the red region? Thanks in advance.
> 
> Jay 

-- 

***************************************************************
* Clemens Vonrhein, Ph.D.     vonrhein AT GlobalPhasing DOT com
*
*  Global Phasing Ltd.
*  Sheraton House, Castle Park 
*  Cambridge CB3 0AX, UK
*--------------------------------------------------------------
* BUSTER Development Group      (http://www.globalphasing.com)
***************************************************************