Configure Continuous Perturbation

The original module for perturbation is for the perturbation of discrete variables only. Continuous variable perturbation is designed for use with continuous variables, particularly those where there may be a number of outlier results such as income. It can operate on both weighted and unweighted datasets.

The continuous perturbation module is also referred to as the "Top Contributors Perturbation Method" and provides specific protection for the following categories of cell:

• Cells where the value is dominated by one or a small number of contributors.
• Cells with a small number of contributors.

If you have continuous variables that are not dominated by a small number of contributors, then you may instead use the perturbed estimates module in conjunction with discrete perturbation. Perturbed estimates simply scales the effect of discrete perturbation proportionately to make sums and means consistent with the perturbed counts.

To use continuous perturbation, you must have:

• R Keys in the unit records.
• An Ftable stored in a CSV file:
  • This file must have exactly 2 columns.
  • The first column represents the rank (with respect to TOPN) and must be an integer value, starting at 1 in the first row and incrementing for each additional row (1,2,3, etc).
  • The second column represents the scaling factor and must be a number (can be a floating point value).
• A CTable stored in a CSV file:
  • This file must have exactly 256 rows and at least 32 columns.
  • All values must be numbers.

Apply the Plugin to a Dataset

The following examples show how to configure continuous perturbation for unweighted and weighted datasets:

Unweighted Dataset

1. Configure the method:

   CODE

   > method addmethod cont_perturbation_method
   > method cont_perturbation_method adddcplugin perturbation continuousperturbationmodule
   > method cont_perturbation_method perturbation addproperty RKEY "true"
   > method cont_perturbation_method perturbation addproperty FREQ "true"
   > method cont_perturbation_method perturbation addproperty TOPN "2"
   > method cont_perturbation_method perturbation addproperty TOPN_RKEY "2"
   > method cont_perturbation_method perturbation addproperty SMALLC "5"
   > method cont_perturbation_method perturbation addproperty "FTABLE" "C:\perturbation\ftable.csv"
   > method cont_perturbation_method perturbation addproperty "CTABLE" "C:\perturbation\ctable.csv"

   

   Notes:

   • TOPN_RKEY is independent from TOPN but must be set to the same value. SuperSERVER processes a list of measure and R Key pairs, ranks the R Keys based on the descending order of the associated measure, and then picks up the top n R Key.
   • The SMALLC property is optional. It defaults to 5 if not specified.
   • If you do not set the FTABLE or CTABLE properties, then they default to a file located in the same location as the SXV4 file, with the extension .ftable or .ctable.

2. Assign the method to a dataset (in this example we are assigning the method to a dataset with the ID bank):

   CODE

   > cat bank addmethod cont_perturbation_method

Weighted Dataset

1. Configure the method:

   CODE

   > method addmethod cont_perturbation_method
   > method cont_perturbation_method adddcplugin perturbation continuousperturbationmodule
   > method cont_perturbation_method perturbation addproperty RKEY "true"
   > method cont_perturbation_method perturbation addproperty FREQ "true"
   > method cont_perturbation_method perturbation addproperty TOPN "2"
   > method cont_perturbation_method perturbation addproperty TOPN_RKEY "2"
   > method cont_perturbation_method perturbation addproperty TOPN_MAIN_WEIGHT "2"
   > method cont_perturbation_method perturbation addproperty SMALLC "5"
   > method cont_perturbation_method perturbation addproperty "FTABLE" "C:\perturbation\ftable.csv"
   > method cont_perturbation_method perturbation addproperty "CTABLE" "C:\perturbation\ctable.csv"

   

   Notes:

   • TOPN_RKEY and TOPN_MAIN_WEIGHT must be set to the same value as TOPN.
   • The SMALLC property is optional. It defaults to 5 if not specified.
   • If you do not set the FTABLE or CTABLE properties, then they default to a file located in the same location as the SXV4 file, with the extension .ftable or .ctable.

2. Assign the method to a dataset (in this example we are assigning the method to a dataset with the ID bank):

   CODE

   > cat bank addmethod cont_perturbation_method