significant figures in reporting Rfree
![](https://secure.gravatar.com/avatar/5b304fab2ec486c5bf68ebdd42821410.jpg?s=120&d=mm&r=g)
Hi. I am wondering what is the common practice with respect to how many significant figures should be used for reporting Rfree at different resolution? Should we always use all 4 figures that are reported by the program? Thank you, Jenn
![](https://secure.gravatar.com/avatar/7984547b53fca6ae837ba286f9c92e29.jpg?s=120&d=mm&r=g)
Hi Jenn,
Hi. I am wondering what is the common practice with respect to how many significant figures should be used for reporting Rfree at different resolution? Should we always use all 4 figures that are reported by the program?
example: 23.11% or 0.2311. I am not aware of this being a function of data quality (resolution or completeness, for example). Pavel. P.S.: In general, programs (should) report values the way you can use them right away without doing any extra work converting them to something else.
![](https://secure.gravatar.com/avatar/1dea6e0a6fdffa6fd747a13f5f5e24fd.jpg?s=120&d=mm&r=g)
On Mon, 2011-07-11 at 14:15 -0400, Jennifer Weinreich wrote:
what is the common practice with respect to how many significant figures should be used for reporting Rfree
If you look up crystallographic papers, you'll find that people routinely report R-values with 3 significant digits, i.e. 23.4% and such. The fourth digit is unreliable - try refining against the same dataset with different test set and you'll find that the Rfree varies within at least 0.1% (depending on resolution, data quality, etc). Most programs report 4 digits, which in my view is needed so that you can properly round it. PDB lists three significant digits as well. -- "I'd jump in myself, if I weren't so good at whistling." Julian, King of Lemurs
![](https://secure.gravatar.com/avatar/7984547b53fca6ae837ba286f9c92e29.jpg?s=120&d=mm&r=g)
what is the common practice with respect to how many significant figures should be used for reporting Rfree If you look up crystallographic papers, you'll find that people routinely report R-values with 3 significant digits, i.e. 23.4% and such. The fourth digit is unreliable - try refining against the same dataset with different test set and you'll find that the Rfree varies within at least 0.1% (depending on resolution, data quality, etc).
You don't even need to change the test set to see that even the third one is not reliable too: pages 75-77 here: http://www.phenix-online.org/presentations/latest/pavel_refinement_general.p... Pavel.
![](https://secure.gravatar.com/avatar/bf0093746a9bfaac88239d39ffff7391.jpg?s=120&d=mm&r=g)
It should probably have some relation to what you are going to put in REMARK 3 of your pdb file when you deposit it: REMARK 3 FREE R VALUE TEST SET SIZE (%) : 1.700 REMARK 3 FREE R VALUE TEST SET COUNT : 4924 REMARK 3 ESTIMATED ERROR OF FREE R VALUE : 0.004 There was a previous thread on how to calculate this value, I think maybe CNS PDB_DEPOSIT script gives it for you. Jennifer Weinreich wrote:
Hi. I am wondering what is the common practice with respect to how many significant figures should be used for reporting Rfree at different resolution? Should we always use all 4 figures that are reported by the program?
Thank you,
Jenn
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
participants (4)
-
Ed Pozharski
-
Edward A. Berry
-
Jennifer Weinreich
-
Pavel Afonine