Updated: April 2005

Index Return to Main Contents

In the first step, the scaling is a simple linear least-squares minimization employing one scaling parameter per scaling group. It then performs an outlier rejection on the scaled data. Subsequently it scales the data using a somewhat more sophisticated scaling algorithm, in which the single-parameter scale parameters are combined with one set of spherical-harmonic parameters for each data run. It then does another round of outlier rejection, restoring to the list of possibly included observations the observations that had been deleted on the previous pass. A third cycle of scaling is then performed with spherical harmonics and with incorporation of nonlinear scaling algorithms if the linear algorithm becomes unstable. If the redundancy in the data is sufficiently high, a fourth round of scaling is performed with the Friedel-mates separated.

After these rounds of scaling, a set of statistical indicia are computed that characterize the data, and the spacegroup of the data is revalidated. At that point the scaled data have been massaged adequately. Therefore the data are written out in six different ASCII formats:

- 1. the traditional X-GEN MULIST format; 2. a format suitable for input to SHELX and XPREP; 3. a format suitable for input to CNS; 4. a format suitable for input to XPLOR v.3.1. 5. a format compatible with the output.sca format from SCALEPACK,
- with Friedel-mates kept separate.
- 6. a format compatible with the output.sca format from SCALEPACK,
- with Friedel-mates merged.

Accordingly the standard behavior of scalem is equivalent to running running the individual X-GEN post-processing programs "scalek", "reject", "stats", "spaceg", and "makemu" in the following sequence:

- 1. scalek -ktp1 2. reject -bf0.015 3. scalek -ktgp1 4. reject -bf0.015 5. scalek -ktghp1 6. scalek -ktghnp1 7. stats -n 8. spaceg 9. makemu -a

If the redundancy is low, the sixth step is removed, and the and the "stats" step will not have the -n flag set.

Most of these Booleans and switches for "scalem" correspond to options in the individual X-GEN post-processing programs, and the descriptions in the "parent" documents are in some cases more complete than those given here.

- -c
- If specified, "scalem" will open a file with environment variable name CHECKREFL each time that it does an outlier rejection and will write out a moderately detailed log of how the rejection is done to that file. This is equivalent to the -c flag to "reject".
- -f
- In the final output files, write out structure amplitudes rather than intensities. This is the equivalent of the -f flag to "makemu".
- -k
- During outlier rejection, if an observation is flagged for a reflection with exactly two observation, then delete both observations. This is the equivalent of the -k flag in "reject".
- -m
- If specified, "scalem" will only write out multiply-observed reflections. This is the equivalent of the -m flag to "makemu".
- -n
- If specified, "scalem" will set the intensities of reflections for which <I> is less than zero to zero just before it writes them out. This is the equivalent of the -n flag to "makemu".
- -o
- If specified, "scalem" will refrain from writing out a list of missing reflections. Otherwise it will print out a list of all reflections that are not contained in the input MULTIREF but which lie within the resolution range associated with that MULTIREF. This is equivalent to the -o option to "makemu".
- -p
- If specified, "scalem" will preserve the sin(theta) dependence of the input scale factors rather than correcting for it. This is equivalent to the -p option to "makemu".
- -r
- During scaling, merge the top and bottom of the detector together rather than calculating scale factors for those two halves separately. This is equivalent to the -t option to "scalek".
- -t
- If specified, "scalem" will output the ASCII results in sin(theta)/lambda order rather than in hkl order. This is equivalent to the -t option to "makemu".
- -u
- If specified, "scalem" will compute the variances of the merged reflections according to an alternate approach described in the "makemu" documentation. This option is equivalent to the "-v" option to "makemu".
- -v
- If specified, "scalem" will print out substantially more information to the log file during each stage of the post- processing effort than would otherwise happen. This is equivalent to the "-v" flags to "scalek", "reject", "stats", and "spaceg".
- -w
- If specified, the initial scaling effort will be based on a Wilson-statistics analysis of the input data rather than by setting the starting scale factors to 1. This is equivalent to the -w flag to "scalek".
- -z
- During outlier rejection, inflate the observation sigmas by a machine-error term. This is equivalent to the -m flag in "reject".
- -a<val>
- This specifies the minimum average redundancy for the
- dataset as a whole, below which the Friedel-separated scaling step will be skipped. The default value of <val> is 4.
- -b<val>
- Specifies the number of resolution bins into which the
- data will be broken for calculations of completeness and redundancy and for shell R-factor calculations during the statistics calculations. This is equivalent to the "-b" flag to "stats". The default is 10 for datasets with more than 27000 observations; it ranges between 4 and 9 for smaller datasets.
- -g<val>
- During outlier rejection, we will attempt to delete a
- fraction of the data equal to <val>. This is equivalent to the "-f" flag to "reject". Default value: 0.015.
- -j<val>
- This specifies the minimum <I> value for a reflection such
- that it can be included in the scaling process. It is therefore equivalent to the "-i" option in "scalek".
- -q<val>
- This specifies the value of |I - <I>| / sigma used during
- outlier rejection to determine whether an observation should be considered for deletion. It is therefore equivalent to the "-q<val>" option to "reject".
- -s<val>
- This specifies an overall scale factor for the ASCII
- output files. It is thus equivalent to the /-s/ option to makemu.

- scalem:
- Scale, reject outliers, and produce ASCII output from a MULTIREF.
- scalem -tvg0.01
- Perform post-processing with the ASCII output in sin(theta)/lambda order, verbose log files, deleting roughly 1% of the data during outlier-rejection.

- NAME
- SYNOPSIS
- DESCRIPTION
- EXAMPLES
- REPORTING BUGS
- COPYRIGHT

This document was created by man2html, using the manual pages.

Time: 02:08:09 GMT, October 03, 2005