Session 4: Statistical Disclosure Limitation Techniques

1
Session 4 Statistical Disclosure Limitation
Techniques

• Julia Lane

2
Overview

• Review of Basic Regression Analysis
• Statistical Disclosure Limitation (SDL) and
  Variable Deletion
• SDL and Variable Perturbation

3
Review of Basic Regression Analysis

• Inference problem
• True relationship is
• Need to infer from data gt

4
Data look as follows
5
OLS minimize sum of squared error
6
Reminder of calculation of regression
coefficients and the standard error of slope

7
Reminder of Assumptions of OLS Regression

• Model is linear in parameters
• The residuals are normally distributed
• The residuals have constant variance
• The expected value of the residuals is always
  zero
• The residuals are independent from one another
• The X values are precise
• The independent variables are not too strongly
  collinear

8
SDL Reduce Information Discussion of impact on
regression coefficients and hypothesis testing

• variable deletion
• recoding categorical variables into larger
  categories
• recoding continuous variables into categories
• rounding continuous variables
• using top and bottom code
• using local suppression and enlarging geographic
  areas

9
SDL Perturb Information Discussion of impact on
regression coefficients and hypothesis testing

• noise addition
• record swapping
• rank swapping
• blanking and imputation
• micro-aggregation
• multiple imputation/modeling to generate
  synthetic data