 |
 |
It seems that a test set was chosen for each set and several selections were combined the wrong way. Just take the merged reflection file, choose a new test and refine the model until R-free and R level out. It could be worthwhile to do a round of manual rebuilding, or two. Your work set has changed a lot too. If you want to avoid the issue of test set bias, you could move random reflections from your current restart to the work set, rather than picking a whole new test set at random. Cheers, Robbie Sent from my Windows 10 phone Van: Dale Tronrud<mailto:[log in to unmask]> Verzonden: vrijdag 16 september 2016 20:26 Aan: [log in to unmask]<mailto:[log in to unmask]> Onderwerp: Re: [ccp4bb] help for deposition Since no one else seems to have responded to your query I'll give it a try. Yes, you will have a problem justifying having a free set that covers 43% of your data. Considering that you appear to have about 20000 reflections you should be able to get by with the default test set size of 5%. 10% is the biggest test set size that I've seen used routinely. A rule-of-thumb is to aim for at least 1000 reflections in your test set, as long as the percentage size doesn't get too large. I don't know of any software that will mark 40% of data with a free flag. Something went terribly wrong when you performed this step. You should go back and verify each step of your data reduction. I think you will find that your refinement will go better when you actually use more of your data. Dale Tronrud On 9/15/2016 7:02 AM, Rakesh Chatterjee wrote: > HI ALL > > i just solved a structure with anisotropy. it has problem of > super-lattice formation and on validation says. the reflection file is > merged from different data sets to gain better completeness. > > *Structure factors* > > Warning: (data block=1): too many reflections for the free set(n free=9676, 43.01%). > > is that going to pose a problem in deposition. i am pretty novice in > this area so if some one has suggestion > > thanks in advance > > -- > Rakesh Chatterjee > Dr Saumen Datta > CSIR-IICB > KOL32