mpxsmooth utility
The mpxsmooth utility is used during the weight generation process.
Previously when weights were generated by the weight generation process, there was almost always a need for manually editing the weights to make them optimal for the implementation. This process is called weight "smoothing". Smoothing is accomplished by reading the matched and unmatched sampling binary data generated by the mpxcomp utility. The utility reads this binary data and applies any necessary adjustments to ensure a smoothed output. The data is considered smoothed when the sample data is monotonically decreasing for msamps (matched samples) and increasing for usamps (unmatched samples). This output can then be used with the mpxwgts utility to generate the smoothed weights used by the operational server. The smoothing process applies only to data that represents one-dimensional (1DIM), two-dimensional (2DIM), and three-dimentional (3DIM) weights.
The reason the sampling data files are generally non-monotonically increasing and decreasing is because of irregular counts in the unmatched and matched samples where they are not expected. Fixing the discrepancy before running the weight generation step where the final weights are computed ensures smoothed weights. The process begins by setting all input values to wgtFLR if they are less than this value. Next, each value is compared with its nearest neighbor to determine if an adjustment must be made. If an adjustment is needed, the slope (derivative) and the mid point of these two values are calculated. The result of these calculations is used to correct the current and next sampling data value. This process is repeated until the data monotonically increases or decreases.
The mpxsmooth utility eliminates
the need for manual intervention. The weight generation utility, which
is started from InfoSphere®
MDM Workbench or
by using the madconfig generate_weights
target, automatically
runs this utility. The only time you might run mpxsmooth is
when you already have the matched sample and unmatched sample binary
files from a previous weight generation and want to smooth them. Manually
running mpxsmooth requires that you also run the mpxwgts utility.
If
the bxmOutDir
option is not provided, then a view
of the before and after smoothing process is dumped to stdout (console).
Both Usamp
and Msamp
options can
be provided at the same time. Remember that this utility replaces
the original binary and text files if the bxmOutDir
setting
is the same as bxmInpDir
.
Before you run a utility, make sure that you have set the necessary operational server environment variables. For information about the variables, see the operational server environment variables topic.
Options and targets | Description | Default |
---|---|---|
entType |
Identifies the entity type for which the sampling data applies. | |
bxmInpDir |
This setting is the directory that contains the matched (msamp) and unmatched (usamp) binary files. | |
bxmOutDir |
This setting is the directory you want the smoothed matched and unmatched files written to. | |
slopeReduction |
This parameter specifies the slope rate of change between data points. The slope calculation is inversely proportional to this value. This means that larger values result in smaller rates of change. The default value is 3. | 3 |
Usamp |
This option directs the utility to smooth the unmatched samples file. | |
Msamp |
This option directs the utility to smooth the matched samples file. |