Run|Sample Balancing: Variable data

Note: Run|Sample Balancing cannot be selected unless your job and data files are open.

Add filter (optional)

Use add filter to weighspefific respondents.
First add a filter i.e. REGION (1)

Add additional filters as needed i.e. REGION (2), REGION (3), REGION (4)

Once all desired filters have been added select the first filter.

Next, choose the desired variables to weight and enter the desired target percents (see below).

Weights will only be applied to the desired filter.

This works great to weight the data based off different target percents by region or other variables.

Choose one or more variables to weight

Move up to 10 variables from the Variable List on the left into the Variables to Weight list box on the right using the right-facing arrow.
Variables defined in the Glossary using the DECLARE instruction are included in the Variable List if the Use glossary transformations option is enabled.
Variables defined in the Glossary using the DEFINE instruction with simple logic, (for example, DEFINE NEW_VAR = 1/75) are included in the Variable List if the Use glossary transformations option is enabled.
Use the left-facing arrow to remove variables from the Variables to Weight list box.

Find a variable

Allows you to enter search criteria to locate a specific variable. This is especially useful when there are many variables in your job file and you want to quickly locate a specific variable.

Find Next

Allows you to find the next variable that matches the previously entered search criteria. This is especially useful when there are multiple variables in your job file with similar variable names.

Set the target percents for each code value for the selected variable

The following information about the selected variable is displayed:

Sequence #

Displays the sequence # for the selected variable.

Code Value(s)

Displays the Code Value(s) found in the data file for the selected variable.

Value Label

Displays the Value Label(s) found in the data file for the selected variable. For variable data files with no Value Label(s), the Value Label displayed will be the same as the Code Value.

Observed Frequency

Displays the frequency for each code value of the selected variable.

Observed Percent

Displays the percent of total for each code value of the selected variable.

Target Percent

Enter the Target Percent for each Code Value or combined levels of Code Value(s).
The Total of all target percents must add to 100%.
There is a right-click context menu to allow Cut, Copy, Paste, Delete and Select All operations in the Target Percent column.

The Use goodness-of-fit minimization technique option of Sample Balancing allows up to 10 variables, with up to 20 numeric Code values (levels) per variable. The total number of combinations cannot exceed 50 million. You can calculate the number of combinations by multiplying the number of code values for the 1st variable by the number of code values for the 2nd variable by the number of code values for the 3rd variable, etc. For example, if you are working with variables that have 6 code values each, up to 9 of these variables can be analyzed at the same time, because 6 to the 9th power is less than 50 million. The Use iterative proportional fitting technique option of Sample Balancing allows up to 50 variables, with up to 100 Code values (levels) per variable.

Clear Percents

Use Clear Percents to initialize Target Percent to 0.00 for all code values of the selected variable.

Combine Levels

Use Combine Levels to combine two or more Code Values into one.

Dissolve Level

Dissolve Level allows you to dissolve a previously combined group of Code Values.

Set Value Label

Highlight anywhere on the row for the desired code value and choose Set Value Label to edit the Value Label for the selected Code Value.
Enter the new Value Label text on the Value Label Text dialog.

Options

Variable to which weight will be written

Choose the variable from the dropdown list of variables in your data file where the weight will be written.
WinCross defaults this variable to the first variable with "WEIGHT" as part of the variable name.
If there is no variable with "WEIGHT" as part of the variable name or if you want to use a different variable in your data file, you will need to Choose a weight variable from the dropdown list of variables or select Create a weight variable to add a new variable to the end of the data file.
To Create a weight variable that will be added to the end of the data file, enter the Name of the new variable and the Width of variable for the new weight variable.
In all cases, the Variable to which the weight will be written, must be at least 5 characters in length.
If you are adding length and defining new variables in the glossary, the length and new variables are not written to the data file from Sample Balancing and variables added in the Glossary cannot be used as the Variable to which weight will be written.

Calculate to filtered sample size

This option is the WinCross default and calculates the weights to the filtered sample size.
Filtered sample size is the total after applying the run filter (if applicable).

Calculate to reduced sample size

Calculates the weights to the reduced sample size.
Reduced sample size is the total of all non-zero target percents.

Calculate to a custom sample size

Calculates the weights to the total entered in Custom sample size.
When using Calculate to a custom sample size, only the weights are adjusted. The Data file sample size, Filtered sample size and Reduced sample size remain the same.

Goodness of fit

Specifies the threshold value for criterion of goodness of fit.

Maximum passes

Specifies the maximum number of passes to take to reach the Goodness of fit threshold.

Use goodness-of-fit minimization technique

Computes the sample balancing using the goodness-of-fit minimization technique and may result in negative weights.
If the goodness-of-fit minimization technique results in negative weights and you have not selected the Replace negative weights with option, an informative message box displays with alternatives to using negative weights.

Replace negative weights with

Replaces any negative weights with the number entered in this field.
The default value for Replace negative weights with is 0.001.

Use iterative proportional fitting technique

Computes the sample balancing using the iterative proportional fitting technique and will never result in negative weights.

Note: The Iterative Proportional Fitting technique never produces negative weights.

Use glossary transformations

Applies glossary statements written in the Setup Glossary Variables dialog prior to the calculation for sample balancing.

Run Filter

Enter a Filter Title (optional) to be applied to your sample balancing calculation.
Enter Filter Logic to be applied to your sample balancing calculation.
You can select Write all cases to file to write all cases to the output data file; however, sample balancing will only be applied to those cases matching the filter criteria.
If you write all cases to the output data file you can check the box Fill cases that do not match filter with '1'. This allows you to assign a weight of 1.0 to those cases that do not meet the filter criteria but are still being written to the output data file.
The Run Filter is active check box can be used to remove the Run Filter without removing the Filter Title and Filter Logic that was previously entered. This allows you to reuse the Filter Title and Filter Logic if desired.

Run

Select Run to write the weights to the open data file or a new data file.
Drive/Directory - Select the file path where you would like to save your new data file with the weights added.
File(s) - If you would like to save your file to a filename that currently exists and is displayed in the File(s) list, highlight the filename from the File(s) list.
Output file - Enter a name for your output data file or select an existing file from the File(s) list.
You can save to the open data file.
You will be prompted with a message explaining that the original data file will be backed up with a *.bak# extension added (with # = to a number, for example, example.sav.bak1 is the first backup file created for the original data file of example.sav).
The newly created data file will automatically reopen in WinCross.
You have the option of continuing by selecting Yes or cancelling the save by selecting No.
Save file as type - Select the file type for your output data file. The output data file must be the same type as the input data file.
Cases to write - Enter the number of cases to write to the output data file.
Select OK to run the sample balancing.
Saves the sample balancing specifications as jobname.sbv to the folder where the open job file currently resides.
If jobname.sbv already exists, it will automatically be replaced.

When Run is selected on the Run Sample Balancing dialog, the Sample Balancing report becomes the active dialog. You can print this report by selecting File|Print|Print Report or you can save the report using File|Save|Save Report or File|Save|Save Report As. The Sample Balancing report can be saved as a Plain Text Report (*.rpt), Microsoft Word/Rich Text Format (*.rtf) report or as a Microsoft Excel 1997-2003 (*.xls) or Microsoft Excel 2007-2013 (*.xlsx) report.

Save

Saves the sample balancing specifications as jobname.sbv to the folder where the open job file currently resides.

Load

Use Load to open a sample balancing specifications file (*.sbv) choosing a different folder/file path from where the open job file currently resides.

Save As

Allows you to save the sample balancing specifications to a file name (*.sbv) and location of your choosing.

Cancel

Use Cancel to exit the Run Sample Balancing dialog. You will be prompted to save any changes made to the Run Sample Balancing dialog.

Restore Defaults

Resets all sample balancing settings to the WinCross defaults.