ReduceX: Suppressed tabular data: Reduce dummy matrix, X (and estimate...

In RegSDC: Information Preserving Regression-Based Tools for Statistical Disclosure Control

Suppressed tabular data: Reduce dummy matrix, X (and estimate Y)

Description

In section 7 in the paper Z = t(X) %*% Y where X is a dummy matrix. Some elements of Y can be found directly as elements in Z. Corresponding rows of X will be removed. After removing rows, some columns will only have zeros and these will also be removed.

Usage

ReduceX(x, z = NULL, y = NULL, digits = 9)

Arguments

x	X as a matrix
z	Z as a matrix
y	Y as a matrix
digits	When non-NULL and when NULL y input, output y estimates close to whole numbers will be rounded using digits as input to RoundWhole.

Details

To estimate Y, this function finds some values directly from Z and other values by running Z2Yhat on reduced versions of X and Z.

Value

A list of four elements:

x	Reduced x
z	Corresponding reduced z or NULL when no z in input
yKnown	Logical vector specifying elements of y that can be found directly as elements in z
y	As y in input (not reduced) or estimated y when NULL y in input

Author(s)

Øyvind Langsrud

Examples

# Same data as in the paper
z <- RegSDCdata("sec7z")
x <- RegSDCdata("sec7x")
y <- RegSDCdata("sec7y")  # Now z is t(x) %*% y 

a <- ReduceX(x, z, y)
b <- ReduceX(x, z)
d <- ReduceX(x, z = NULL, y)  # No z in output

# Identical output for x and z
identical(a$x, b$x)
identical(a$x, d$x)
identical(a$z, b$z)

# Same y in output as input
identical(a$y, y)
identical(d$y, y)

# Estimate of y (yHat) when NULL y input
b$y

# These elements of y can be found directly in in z
y[a$yKnown, , drop = FALSE]
# They can be found by searching for unit colSums
colSums(x)[colSums(x) == 1]

# These trivial data rows can be omitted when processing data
x[!a$yKnown, ]
# Now several columns can be omitted since zero colSums
colSums0 <- colSums(x[!a$yKnown, ]) == 0
# The resulting matrix is output from the function
identical(x[!a$yKnown, !colSums0], a$x)

# Output z can be computed from this output x
identical(t(a$x) %*% y[!a$yKnown, , drop = FALSE], a$z)

RegSDC documentation built on Aug. 19, 2022, 9:08 a.m.