There are Rpim and Rrim, Rpim is sqrt(1/(N-1)) and is usually small and Rrim
(or Rmeas)=sqrt(N/(N-1)) and is large. I usually go with I/sigma cutoff.
Maia
----- Original Message -----
From: "Edward A. Berry" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, April 22, 2010 11:59 AM
Subject: Re: [ccp4bb] Rsym problems...maybe???
> There are plenty of structures in the database with R-sym=0.99.
> But something is odd here. If I understand R-pim, it should
> always be bigger than Rsym, because this factor of sqrt(N/(N-1)) is always
> >1
> Are you saying Rpim is .30 and Rsym is 1.00?
>
> Last time I deposited a structure, Rsym and Rmerge in the last shell are
> optional.
> I would leave it out and rely on the excellent I/sigI in the last shell,
> and use all the data (provided after refinement R-free in the last shell
> is < .4).
> Ed
>
> Daniel Bonsor wrote:
>> Hello again.
>>
>> At first I was not worry but maybe now I am. I have completed a structure
>> and submitted to the PDB. They queried my Rsym value in the highest
>> resolution bin, 2.5-2.37A (may I dare say it 100%). I was not worried at
>> the time as I had:
>>
>> 99.4% completeness
>> Mean(I/sdI) of 2.5
>> and a redundancy of 11 (which would explain the high Rsym)
>> Space group I422
>>
>> My Rpim in this shell is 30%.
>>
>> Should I reduce the resolution and start from scratch again or is
>> everything fine and dandy and I should stop worrying?
>>
>
|