GaussSuppressDec: Cell suppression with synthetic decimal numbers

In GaussSuppression: Tabular Data Suppression using Gaussian Elimination

Cell suppression with synthetic decimal numbers

Description

GaussSuppressionFromData, or one of its wrappers, is run and decimal numbers are added to output by executing SuppressDec.

Usage

GaussSuppressDec(
  data,
  ...,
  fun = GaussSuppressionFromData,
  output = NULL,
  use_freqVar = NA,
  digits = 9,
  nRep = NULL,
  rmse = pi/3,
  sparseLimit = 500,
  rndSeed = 123,
  runIpf = FALSE,
  eps = 0.01,
  iter = 100,
  mismatchWarning = TRUE,
  whenDuplicatedInner = NULL,
  whenMixedDuplicatedInner = warning
)

Arguments

data	Input daata as a data frame
...	Further parameters to GaussSuppressionFromData
fun	A function: GaussSuppressionFromData or one of its wrappers such as SuppressSmallCounts and SuppressDominantCells.
output	NULL (default), "publish", "inner", "publish_inner", or "publish_inner_x" (x also).
use_freqVar	Logical (TRUE/FALSE) with a default value of NA. Determines whether the variable freqVar is used as the basis for generating decimal numbers. If NA, the parameter is set to TRUE, except in the following cases, where it is set to FALSE: If freqVar is not available. If runIpf is FALSE and fun is one of the functions SuppressFewContributors or SuppressDominantCells. When use_freqVar is FALSE, only zeros are used instead. This approach is more robust in practice, as decimal numbers can then be stored more accurately. The default value is chosen to ensure compatibility with existing code and to allow for the use of freqVar when dealing with frequency tables, which may be useful.
digits	Parameter to RoundWhole. Values close to whole numbers will be rounded.
nRep	NULL or an integer. When >1, several decimal numbers will be generated.
rmse	Desired root mean square error of decimal numbers. Variability around the expected, according to the linear model, inner frequencies. The expected frequencies are calculated from the non-suppressed publishable frequencies.
sparseLimit	Limit for the number of rows of a reduced x-matrix within the algorithm. When exceeded, a new sparse algorithm is used.
rndSeed	If non-NULL, a random generator seed to be used locally within the function without affecting the random value stream in R.
runIpf	When TRUE, additional frequencies are generated by iterative proportional fitting using Mipf.
eps	Parameter to Mipf.
iter	Parameter to Mipf.
mismatchWarning	Whether to produce the warning "⁠Mismatch between whole numbers and suppression⁠", when relevant. When nRep>1, all replicates must satisfy the whole number requirement for non-suppressed cells. When mismatchWarning is integer (⁠>0⁠), this will be used as parameter digits to RoundWhole when doing mismatch checking (can be quite low when nRep>1).
whenDuplicatedInner	Function to be called when default output and when cells marked as inner correspond to several input cells (aggregated) since they correspond to published cells.
whenMixedDuplicatedInner	Function to be called in the case above when some inner cells correspond to published cells (aggregated) and some not (not aggregated).

Value

A data frame where inner cells and cells to be published are combined or output according to parameter output.

Author(s)

Øyvind Langrsud

See Also

SuppressionFromDecimals()

Examples

a <- GaussSuppressDec(data = SSBtoolsData("example1"), 
                      fun = SuppressSmallCounts, 
                      dimVar = c("age", "geo"),
                      preAggregate = TRUE, 
                      freqVar = "freq", maxN = 3)
a                       
                 

b <- GaussSuppressDec(data = SSBtoolsData("magnitude1"), 
                      fun = SuppressDominantCells, 
                      numVar = "value", 
                      formula = ~sector2 * geo + sector4 * eu,
                      contributorVar = "company", k = c(80, 99))
b  
 
# FormulaSelection() works on this output as well 
FormulaSelection(b, ~sector2 * geo)                       
                      

GaussSuppression documentation built on April 3, 2025, 7:54 p.m.