Re: [ccp4bb]: A simple question of resolution

[James Stroud <James.Stroud@Colorado.EDU>]

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

Am I confused? Where is my logic flawed here?

If an error model gives chi2 of ~1 across all resolution bins, then we know
that for a typical data set we have a >90% chance that our observations fall
within our error estimates. In essence, we have a fair bit of confidence in
our error model. With this error model, we can now weight our observations
when we fit the physical model (the structure). We can force our chi2 down
during scaling by increasing the error and thus decreasing the weight of the
observation by lowering the I/sigma for the bin. Thus, by having a standard
of chi2 ~1 and I/sigma =2, we ensure that, when computing our physical
model, we do not include a bunch of meaningless observations in bins that
have been weighted to virtual nonexistence. Since our error models make the
assumption of independent uncertainty, we assume that the error increases
only as the square root while the signal increases linearly over a number of
observations. Thus, we assume that we have weak reflections at high
resolution because of random noise and not because a systematic error
overcomes the intensity of the reflection. NMR spectroscopy makes use of
this fact. Thus, redundancy serves a couple of purposes: it helps to refine
the error model and it increases resolution by lowering the error of weak
reflections. So my question is: why is chi2 ~1 and I/sigma =2 (from scaling)
not an acceptable standard if our working R and free R are acceptable at
these resolutions?

James

--------------------------------------------------------
"Research: If it worked the first time,
they would just call it 'search'."     -Roy Garcia
--------------------------------------------------------
 James C. Stroud
 Department of Chemistry and Biochemistry
 University of Colorado at Boulder
 Boulder, CO 80309

 Tel: 303-492-4503      Fax: 303-735-1347
--------------------------------------------------------