Previous: transhkl: Apply a unit cell transformation to a reflection file
Next: scalehkl: Scale intensities by a fixed factor

comparehkl: Compare reflection intensities

The "comparehkl" program compares reflections within a reflection file (resulting in a R-merge or R-int values) and between reflection files (giving a scale factor and R-int values).

Usage

comparehkl [options] [.x or .rmat file] <one or two .hkl or .sca files>

Accepted options

The program accepts all options for controlling the way the unit cell is treated, plus:

shells=#: Print statistics about different shells of data, give the desired number of shells as the argument (e.g. shells=10)
dmax=#.## and dmin=#.##: Limit the statistics to data that are within the specified resolution limits. Resolution limits should be given in Angstrom. Either specify neither or both, if only one is specified, it is ignored.

Description of the output

Intensity

Intensity   : <I>= 133.7; <sig>=   6.4; <I>/<sig>=  20.9

The average intensity, average sigma and average I/sigma for all reflections.

Redundancy

Redundancy  :   1.3 (10%=  1.0; 50%=  1.0; 90%=  2.1)

The number before the brackets is the average redundancy. Between brackets the 10th percentile, the 50th percentile (median) and the 90th percentile redundancy values are given. In this example, more than 50% of all data is present only once, and 10% of the data "2.1 or more" times.

Completeness

Completeness:  484 out of 1189 ( 40.7%)

Based on the unit cell, the completeness of the data are calculated.

Consistency

Consistency : R1=0.021; R2=0.014; Rmeas=0.030; Chi2=0.87; RmrgdI=0.043

The R1, R2 and Chi2 numbers are similar to the internal consistency measures that scalepack produces (but note that "comparehkl" doesn't do scaling or outlier rejection), Rmeas and RmrgdI are measures of data quality (see next section).

Reflection on R-int

The classical R-int or R-merge is a very bad indicator of data quality. In general, when the redundancy of a data set is higher, the R-merge will be higher. This makes R-merge useless in determining the Laue symmetry of a data set (Imagine an orthorhombic data set that is suspected of being monoclinic. Just the reduction of the redundancy that this implies will already drop R-merge, possibly leading to a wrong conclusion).

There are alternative R factors in the "comparehkl" output that do not suffer from this effect:

R-meas gives a good, stable measurement describing the internal consistency of the input data set. R-meas is always (except at infinite redundancy) higher than R-merge.
R-mrgd gives a pessimistic measurement of the quality of the merged data set (the real quality could be a factor of sqrt(2) better).

For a detailed description of the different merging R-factors and their merits, see Kay Diederichs and P. Andrew Karplus, "Improved R-factors for diffraction data analysis in macromolecular crystallography", Nature Structural Biology Volume 4 Number 4 [April 1997] pages 269--275 and an erratum in a later issue.

Examples

devel[408]ambi%% comparehkl scale_all.sca
{Input cell   : a=7.6567 b=7.8559 c=11.0783 alpha=90.000 beta=90.000 gamma=90.000 P
 Reduced cell : a=7.6567 b=7.8559 c=11.0783 alpha=90.000 beta=90.000 gamma=90.000
 Conventional : a=7.6567 b=7.8559 c=11.0783 alpha=90.000 beta=90.000 gamma=90.000 P
 Volume       : 666.36; System : orthorhombic}
Applied symmetry equivalence:  CentricOrthorhombic
Reading reflection file... WARNING: inverting SCA file hkl/-h-k-l
0.50 seconds
scale_all.sca
=============
All data      : Intensity   : <I>= 133.7; <sig>=   6.4; <I>/<sig>=  20.9
                Redundancy  :   1.3 (10%=  1.0; 50%=  1.0; 90%=  2.1)
                Completeness:  484 out of 1189 ( 40.7%)
                Consistency : R1=0.021; R2=0.014; Rmeas=0.030; Chi2=0.87; RmrgdI=0.043

devel[411]cubic%% comparehkl denzo.hkl shells=10
{Input cell   : a=10.0678 b=10.0649 c=10.0728 alpha=109.468 beta=109.470 gamma=109.461 P
 Reduced cell : a=10.0649 b=10.0678 c=10.0728 alpha=109.470 beta=109.468 gamma=109.461
 Conventional : a=11.6270 b=11.6251 c=11.6284 alpha=89.995 beta=89.952 gamma=90.020 I
 Volume       : 1571.75; System : cubic}
Applied symmetry equivalence:  CentricCubic
Reading reflection file... 17.09 seconds
denzo.hkl
=========
  5.81   1.29 : Intensity   : <I>=2146.3; <sig>=  38.1; <I>/<sig>=  56.3
                Redundancy  :  55.8 (10%= 10.5; 50%= 50.5; 90%= 98.5)
                Completeness:   53 out of   53 (100.0%)
                Consistency : R1=0.077; R2=0.093; Rmeas=0.079; Chi2=3.12; RmrgdI=0.049
  1.29   1.03 : Intensity   : <I>=1123.0; <sig>=   7.9; <I>/<sig>= 143.0
                Redundancy  :  66.1 (10%= 40.0; 50%= 82.0; 90%= 91.0)
                Completeness:   44 out of   44 (100.0%)
                Consistency : R1=0.080; R2=0.112; Rmeas=0.081; Chi2=3.00; RmrgdI=0.028
  1.03   0.90 : Intensity   : <I>= 428.1; <sig>=   3.4; <I>/<sig>= 125.5
                Redundancy  :  60.2 (10%= 36.5; 50%= 49.0; 90%= 87.5)
                Completeness:   42 out of   42 (100.0%)
                Consistency : R1=0.075; R2=0.084; Rmeas=0.075; Chi2=2.23; RmrgdI=0.031
  0.90   0.82 : Intensity   : <I>= 580.0; <sig>=   6.7; <I>/<sig>=  86.9
                Redundancy  :  59.1 (10%= 33.5; 50%= 69.5; 90%= 82.5)
                Completeness:   41 out of   41 (100.0%)
                Consistency : R1=0.081; R2=0.099; Rmeas=0.083; Chi2=1.93; RmrgdI=0.041
  0.82   0.76 : Intensity   : <I>= 388.7; <sig>=   3.6; <I>/<sig>= 109.1
                Redundancy  :  57.8 (10%= 29.5; 50%= 67.5; 90%= 76.0)
                Completeness:   40 out of   40 (100.0%)
                Consistency : R1=0.081; R2=0.092; Rmeas=0.082; Chi2=1.74; RmrgdI=0.022
  0.76   0.71 : Intensity   : <I>= 294.3; <sig>=   4.6; <I>/<sig>=  64.2
                Redundancy  :  50.9 (10%= 26.5; 50%= 60.5; 90%= 70.0)
                Completeness:   41 out of   41 (100.0%)
                Consistency : R1=0.098; R2=0.107; Rmeas=0.099; Chi2=1.56; RmrgdI=0.023
  0.71   0.68 : Intensity   : <I>= 241.6; <sig>=   3.0; <I>/<sig>=  79.4
                Redundancy  :  49.6 (10%= 27.0; 50%= 55.5; 90%= 65.2)
                Completeness:   37 out of   37 (100.0%)
                Consistency : R1=0.100; R2=0.110; Rmeas=0.101; Chi2=1.15; RmrgdI=0.020
  0.68   0.65 : Intensity   : <I>= 348.9; <sig>=   5.5; <I>/<sig>=  63.8
                Redundancy  :  48.7 (10%= 26.0; 50%= 55.5; 90%= 61.5)
                Completeness:   36 out of   36 (100.0%)
                Consistency : R1=0.110; R2=0.113; Rmeas=0.112; Chi2=1.36; RmrgdI=0.045
  0.65   0.62 : Intensity   : <I>= 132.6; <sig>=   2.8; <I>/<sig>=  47.0
                Redundancy  :  41.7 (10%= 21.5; 50%= 49.5; 90%= 58.5)
                Completeness:   39 out of   39 (100.0%)
                Consistency : R1=0.128; R2=0.130; Rmeas=0.130; Chi2=1.18; RmrgdI=0.040
  0.62   0.60 : Intensity   : <I>=  29.7; <sig>=   1.4; <I>/<sig>=  20.6
                Redundancy  :  43.2 (10%= 23.5; 50%= 48.5; 90%= 53.5)
                Completeness:   42 out of   42 (100.0%)
                Consistency : R1=0.236; R2=0.174; Rmeas=0.239; Chi2=0.82; RmrgdI=0.103
All data      : Intensity   : <I>= 627.6; <sig>=   8.7; <I>/<sig>=  72.5
                Redundancy  :  53.6 (10%= 25.5; 50%= 52.6; 90%= 85.5)
                Completeness:  415 out of  415 (100.0%)
                Consistency : R1=0.084; R2=0.100; Rmeas=0.085; Chi2=1.95; RmrgdI=0.038

comparehkl denzo.hkl shells=10  32.54s user 0.44s system 87% cpu 37.519 total

Bugs

"comparehkl" is too slow.

Previous: transhkl: Apply a unit cell transformation to a reflection file
Next: scalehkl: Scale intensities by a fixed factor