IdentificationRiskCalculation

Calculating the Identification Risks in Partially Synthetic Microdata

This package calculates the identification risk in partially synthetic microdata. The expected match risk, the true match rate, and the false match rate are reported. The calculation supports mixed data type, including categorical variables and continuous variables. The details of the calculation methods, including how identification risks are defined, are available at this published paper:

Hornby, R. and Hu, J. (2021), Identification risks evaluation of partially synthetic data with the IdentificationRiskCalculation R package, Transactions on Data Privacy, 14:1, 37-52. http://www.tdp.cat/issues21/abs.a425a21.php

We use the synthpop to synthesize variables.

install.packages("synthpop")
devtools::install_github("https://github.com/RyanHornby/IdentificationRiskCalculation")

require(IdentificationRiskCalculation)
require(synthpop)

Here we provide demonstration on how to use this package. CEdata is available in the package. It contains a few continuous and a few categorical variables.

CEdata <- IdentificationRiskCalculation::CEdata
CEdata$Urban <- as.factor(CEdata$Urban)
CEdata$Marital <- as.factor(CEdata$Marital)
CEdata$Tenure <- as.factor(CEdata$Tenure)

Assume that Age, Urban, Marital are known variables. We use a 0.1 radius for continuous Age variable.

knownvars <- c("Age", "Urban", "Marital")
r_age <- 0.1

For the following 5 scenarios, we use the same 0.1 radius for all continuous variables for demonstration purpose. The output contains the expected match risk exp.risk, the true match rate true.rate, and the false match rate false.rate.

Scenario 1: Income (continuous)

synvars1 <- c("Income")
syndata1 <- syn(CEdata, m = 20, visit.sequence = synvars1)
r_income <- 0.1
riskList1 <- IdentificationRisk(origdata = CEdata, 
                                syndata = syndata1$syn, 
                                known = knownvars, 
                                syn = synvars1, 
                                r = c(r_age, r_income))

exp.risk1 <- riskList1$exp.risk_vector
true.rate1 <- riskList1$true.rate_vector
false.rate1 <- riskList1$false.rate_vector

Scenario 2: Tenure (categorical), Income (continuous)

synvars2 <- c("Tenure","Income")
syndata2 <- syn(CEdata, m = 20, visit.sequence = synvars2)
r_income <- 0.1
riskList2 <- IdentificationRisk(origdata = CEdata, 
                                syndata = syndata2$syn, 
                                known = knownvars, 
                                syn = synvars2, 
                                r = c(r_age, r_income))
                                
exp.risk2 <- riskList2$exp.risk_vector
true.rate2 <- riskList2$true.rate_vector
false.rate2 <- riskList2$false.rate_vector

Scenario 3: Expenditure (continuous), Income (continuous)

synvars3 <- c("Expenditure","Income")
syndata3 <- syn(CEdata, m = 20, visit.sequence = synvars3)
r_income <- 0.1
r_expenditure <- 0.1
riskList3 <- IdentificationRisk(origdata = CEdata, 
                                syndata = syndata3$syn, 
                                known = knownvars,
                                syn = synvars3, 
                                r = c(r_age, r_expenditure, 
                                      r_income))

exp.risk3 <- riskList3$exp.risk_vector
true.rate3 <- riskList3$true.rate_vector
false.rate3 <- riskList3$false.rate_vector

Scenario 4: Tenure (categorical), Expenditure (continuous), Income (continuous)

synvars4 <- c("Tenure", "Expenditure", "Income")
syndata4 <- syn(CEdata, m = 20, visit.sequence = synvars4)
r_income <- 0.1
r_expenditure <- 0.1
riskList4 <- IdentificationRisk(origdata = CEdata, 
                                syndata = syndata4$syn, 
                                known = knownvars, 
                                syn = synvars4,
                                r = c(r_age, r_expenditure, 
                                      r_income))

exp.risk4 <- riskList4$exp.risk_vector
true.rate4 <- riskList4$true.rate_vector
false.rate4 <- riskList4$false.rate_vector

Scenario 5: Tenure (categorical)

synvars5 <- c("Tenure", "Urban")
syndata5 <- syn(CEdata, m = 20, visit.sequence = synvars5)
riskList5 <- IdentificationRisk(origdata = CEdata, 
                                syndata = syndata5$syn, 
                                known = knownvars, 
                                syn = synvars5)

exp.risk5 <- riskList5$exp.risk_vector
true.rate5 <- riskList5$true.rate_vector
false.rate5 <- riskList5$false.rate_vector