[phenixbb] Phenix.refine coordinate error estimate discrepancies

[Phil Jeffrey]

There seems to be a large discrepancy between the likelihood-based 
coordinate error estimate depending on whether the intensities or 
structure factor magnitudes are used, which doesn't make any sense to me.

If I do:

% phenix.refine rebuild.pdb data_withI.mtz

phenix.refine picks up Imean and SigImean from the MTZ and refines the 
structure

     xray_data {
       file_name = "data_withI.mtz"
       labels = "IMEAN,SIGIMEAN"
       r_free_flags {
         file_name = "data_withI.mtz"
         label = "FreeR_flag"
         test_flag_value = 0
       }

then I get:

| maximum likelihood estimate for coordinate error:   1.39 A 



But if I create an MTZ file without the intensities by using CAD on the 
original file, and use the same coordinate file:

% phenix.refine rebuild.pdb data_withoutI.mtz

phenix.refine picks up the structure factors instead:

     xray_data {
       file_name = "data_withoutI.mtz"
       labels = "F,SIGF"
       r_free_flags {
         file_name = "data_withoutI.mtz"
         label = "FreeR_flag"
         test_flag_value = 0
       }
     }

then this seems to be the right ballpark for a 1.75 Angstrom structure....

| maximum likelihood estimate for coordinate error:   0.28 A

Similar R-factors, but there is a factor of 10 difference in the scale 
reported here:

I's:
r_work= 0.2266 r_free= 0.2570 ksol= 0.41 Bsol= 50.00 scale= 0.097
F's:
r_work= 0.2221 r_free= 0.2557 ksol= 0.41 Bsol= 50.00 scale= 0.965


   Version: 1.4
   Release tag: 3
   cctbx tag: 2008_12_07_1353
   Platform: mac-intel-osx osx-10.5.6
   User: phil

Cheers,
Phil Jeffrey
Princeton