NoisyCoverMargins: Laplace or pTable perturbed margins

In olangsrud/experimentalRpackage: An experimental R package containing experimental/early R functions

Description

The function make use of CoverMarginsExact.

Usage

NoisyCoverMargins(
  data,
  freqVar,
  formula,
  formulaExact = NULL,
  eps = 0.5,
  pMatrix = NULL,
  keyVar = NULL,
  ...
)

Arguments

data	Input data as a data frame (inner cells)
freqVar	Variable holding counts (name or number)
formula	Model formula defining cells to be perturbed
formulaExact	Model formula defining cells to be exact
eps	Differential privacy parameter
pMatrix	Output from Pmatrix
keyVar	Variable holding uniformly distributed keys (name or number). See details.
...	Further parameters to CoverMarginsExact

Details

In this function individual keys (cell keys) are not taken as input. Possible individual keys must be "summed up" to inner cell keys beforehand. Zeros keys corresponding to zero frequencies are unproblematic when zeros are not perturbed. By default (keyVar==NULL), the keys corresponding to nonzero frequencies are generated by runif (otherwise zero).

Value

A list

Author(s)

Øyvind Langsrud

Examples

z2 <- EasyData("z2")
a1 <- NoisyCoverMargins(z2, "ant", ~region + kostragr * hovedint)
a2 <- NoisyCoverMargins(z2, "ant", ~region + kostragr * hovedint, ~kostragr + hovedint, eps = 0.25)
a3 <- NoisyCoverMargins(cbind(z2,keyVar = runif(44) * (z2$ant>0)), "ant", 
                        ~region + kostragr * hovedint, ~kostragr + hovedint, 
                        pMatrix = Pmatrix(), keyVar = "keyVar")

olangsrud/experimentalRpackage documentation built on Sept. 20, 2021, 10:32 p.m.