mfisch
I have a disordered oxygen site of which I know from single crystal the positions and that the total occupancy is one. This means that there are several low occupancies on different (close) positions which sum up to one.
With time resolved data, any of these positions can be between one and zero occupied, but again, with a total sum of 1.
I am just not sure whether use
penalty = ((occ_O1 + occ_O2 + occ_O3) -1) ; or
penalty = ((occ_O1 + occ_O2 + occ_O3) -1) ^2; or
restraint= ((occ_O1 + occ_O2 + occ_O3) -1) ; or
restraint= ((occ_O1 + occ_O2 + occ_O3) -1)^2 ;
all of the mentioned possibilities work and give reasonnable solutions. But which one is the best and makes most sense? What about (wrong) local minima?
There are several error calculation methods when using penalties and restraints. Which one to use?
Correlations are high, use_LU_for_errors? Or bootstrap?
johnsoevans
Martin,
Probably not a helpful response, but have you seen the section in tech ref on restraints and penalties (5.5)? For the penalty you need the ^2 to keep things positive.
John