Hi All, Thanks for Phenix! Anyway quick question, I tried out the "generate Table1" utility for fun and noticed that my I/SigI and completeness do not match what is found in my XSCALE file. From Table 1 Completeness (%) 98.62 (90.57) Mean I/sigma(I) 7.39 (1.59) Wilson B-factor 59.78 R-sym R-factor 0.2006 (0.3275) R-free 0.2241 (0.3739) From XSCALE SUBSET OF INTENSITY DATA WITH SIGNAL/NOISE >= -3.0 AS FUNCTION OF RESOLUTION RESOLUTION NUMBER OF REFLECTIONS COMPLETENESS R-FACTOR R-FACTOR COMPARED I/SIGMA R-meas CC(1/2) Anomal SigAno Nano LIMIT OBSERVED UNIQUE POSSIBLE OF DATA observed expected Corr 20.00 313 69 72 95.8% 3.6% 4.6% 313 30.88 4.3% 99.9* 8 0.725 54 10.00 2174 471 471 100.0% 3.9% 4.8% 2174 28.71 4.4% 99.8* 34* 0.972 382 6.00 8195 1983 1992 99.5% 4.7% 5.3% 8160 20.70 5.4% 99.8* 12* 0.913 1299 5.00 8055 1818 1820 99.9% 7.0% 6.9% 8051 17.73 7.9% 99.5* 7 0.892 1337 4.50 7207 1634 1636 99.9% 6.0% 6.2% 7207 20.06 6.8% 99.6* -1 0.824 1163 4.00 9603 2479 2521 98.3% 6.8% 6.7% 9536 17.05 7.9% 99.4* -2 0.819 1369 3.60 13641 3141 3156 99.5% 9.6% 9.5% 13621 13.71 11.0% 99.1* -2 0.799 2073 3.20 22112 4948 4951 99.9% 16.5% 16.6% 22083 8.82 18.8% 98.1* 4 0.810 3383 3.00 14155 3550 3579 99.2% 27.8% 27.7% 14058 5.45 32.1% 94.6* -1 0.742 1875 2.80 19676 4584 4611 99.4% 44.9% 46.2% 19604 3.71 51.2% 88.5* 3 0.740 2750 2.60 26749 6192 6202 99.8% 76.3% 78.4% 26686 2.20 87.2% 71.8* 1 0.717 3598 2.40 33999 8339 8388 99.4% 130.9% 132.7% 33810 1.26 150.6% 48.5* 1 0.671 4170 2.30 22347 5332 5348 99.7% 198.4% 206.4% 22276 0.85 227.7% 31.5* 0 0.623 2685 total 188226 44540 44747 99.5% 9.0% 9.4% 187579 7.09 10.3% 99.8* 2 0.755 26138 How does Phenix compute these parameters? Obviously my I/sigI has gotten much better (1.59 versus 0.85) but the scaling statistics should be the correct ones. Thanks, Drew
On Thu, Oct 18, 2012 at 7:53 AM, Andrew Waight
Thanks for Phenix! Anyway quick question, I tried out the "generate Table1" utility for fun and noticed that my I/SigI and completeness do not match what is found in my XSCALE file. ... How does Phenix compute these parameters? Obviously my I/sigI has gotten much better (1.59 versus 0.85) but the scaling statistics should be the correct ones.
It computes these statistics directly from the reflection data you provide, not the log file. (The reasoning being that since these are the data that Phenix actually uses, the statistics calculated from them are more accurate and relevant than whatever the scaling program thinks.) My guess is that some of the internal processing accounts for the difference - could you please send me the data file? FYI, the plan for the future is to recalculate all of these statistics from unmerged data rather than rely on parsing data processing logfiles, which obey no standard format and are subject to change without notice. -Nat
Sorry, I was not completely paying attention either. One problem here
is that the scaling statistics are divided into 13 bins of uneven
size, whereas phenix.table_one always uses 10 bins, with equal numbers
of possible reflections in each bin (since this seems to be the most
common convention). I should (and will) make the number of bins
adjustable, but clearly the statistics for the top bin in the XSCALE
log will not match what Phenix outputs, regardless of how they're
calculated. This is another reason to move away from reliance on log
files and re-merge the data ourselves.
I'd still like to see the data file, however, as the completeness
calculated by Phenix is significantly lower and can't be accounted for
by the different binning.
thanks,
Nat
On Thu, Oct 18, 2012 at 8:07 AM, Nathaniel Echols
On Thu, Oct 18, 2012 at 7:53 AM, Andrew Waight
wrote: Thanks for Phenix! Anyway quick question, I tried out the "generate Table1" utility for fun and noticed that my I/SigI and completeness do not match what is found in my XSCALE file. ... How does Phenix compute these parameters? Obviously my I/sigI has gotten much better (1.59 versus 0.85) but the scaling statistics should be the correct ones.
It computes these statistics directly from the reflection data you provide, not the log file. (The reasoning being that since these are the data that Phenix actually uses, the statistics calculated from them are more accurate and relevant than whatever the scaling program thinks.) My guess is that some of the internal processing accounts for the difference - could you please send me the data file?
FYI, the plan for the future is to recalculate all of these statistics from unmerged data rather than rely on parsing data processing logfiles, which obey no standard format and are subject to change without notice.
-Nat
participants (2)
-
Andrew Waight
-
Nathaniel Echols