Re: [ccp4bb]: 3Fo-2Fc maps

[Edward Berry <eaberry@lbl.gov>]

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



Stefano Trapani wrote:
> 
> So, can I say (very crudely) that a (3mFo-2DFc) map is better than a
> (2mFo-DFc) map, in that

A very simplistic point comes to mind- 
this (incrementing the coefficients) is going in the direction of 
subtracting large numbers to get a small difference and so will 
increase noise, which at some point must outweigh the advantage 
of lower bias.

Iff we can assume from final R-free values that average total error
(random + systematic) of a structure factor amplitude is around 20%,
when you double Fo and subtract  Fc (~=Fo) the error becomes 40%,
when you triple Fo and subtract 2Fc  the error becomes 60%

so
A. you do not want to make a 10Fo-9Fc map
B. If the weighting factors do a good job of removing bias already in the
(2mFo-DFc) map, then the (3mFo-2DFc) map might be less interpretable due
to increased noise.

I'm sure others will have more quantitative and theoretically 
well-grounded advice on this question.


Ed