Small count rounding of necessary inner cells are performed so that all small frequencies of cross-classifications to be published (publishable cells) are rounded. The publishable cells can be defined from a model formula, hierarchies or automatically from data.
Usage
PLSrounding(
data,
freqVar = NULL,
roundBase = 3,
hierarchies = NULL,
formula = NULL,
dimVar = NULL,
maxRound = roundBase - 1,
printInc = nrow(data) > 1000,
output = NULL,
extend0 = FALSE,
preAggregate = is.null(freqVar),
aggregatePackage = "base",
aggregateNA = TRUE,
aggregateBaseOrder = FALSE,
rowGroupsPackage = aggregatePackage,
...
)
PLSroundingInner(..., output = "inner")
PLSroundingPublish(..., output = "publish")Arguments
- data
Input data (inner cells), typically a data frame, tibble, or data.table. If
datais not a classic data frame, it will be coerced to one internally unlesspreAggregateisTRUEandaggregatePackageis"data.table".- freqVar
Variable holding counts (inner cells frequencies). When
NULL(default), microdata is assumed.- roundBase
Rounding base
- hierarchies
List of hierarchies
- formula
Model formula defining publishable cells
- dimVar
The main dimensional variables and additional aggregating variables. This parameter can be useful when hierarchies and formula are unspecified.
- maxRound
Inner cells contributing to original publishable cells equal to or less than maxRound will be rounded
- printInc
Printing iteration information to console when TRUE
- output
Possible non-NULL values are
"input","inner"and"publish". Then a single data frame is returned.- extend0
When
extend0is set toTRUE, the data is automatically extended. This is relevant whenzeroCandidates = TRUE(seeRoundViaDummy). Additionally,extend0can be specified as a list, representing thevarGroupsparameter in theExtend0function. Can also be set to"all"which means that input codes in hierarchies are considered in addition to those in data.- preAggregate
When
TRUE, the data will be aggregated beforehand within the function by the dimensional variables.- aggregatePackage
Package used to preAggregate. Parameter
pkgtoaggregate_by_pkg.- aggregateNA
Whether to include NAs in the grouping variables while preAggregate. Parameter
include_natoaggregate_by_pkg.- aggregateBaseOrder
Parameter
base_ordertoaggregate_by_pkg, used when preAggregate. The default is set toFALSEto avoid unnecessary sorting operations. WhenTRUE, an attempt is made to return the same result withdata.tableas with base R. This cannot be guaranteed due to potential variations in sorting behavior across different systems.- rowGroupsPackage
Parameter
pkgtoRowGroups. The parameter is input toFormula2ModelMatrixviaModelMatrix.- ...
Further parameters sent to
RoundViaDummy
Value
Output is a four-element list with class attribute "PLSrounded",
which ensures informative printing and enables the use of FormulaSelection on this object.
- inner
Data frame corresponding to input data with the main dimensional variables and with cell frequencies (original, rounded, difference).
- publish
Data frame of publishable data with the main dimensional variables and with cell frequencies (original, rounded, difference).
- metrics
A named character vector of various statistics calculated from the two output data frames ("
inner_" used to distinguish). See examples below and the functionHDutility.- freqTable
Matrix of frequencies of cell frequencies and absolute differences. For example, row "
rounded" and column "inn.4+" is the number of rounded inner cell frequencies greater than or equal to4.
Details
This function is a user-friendly wrapper for RoundViaDummy with data frame output and with computed summary of the results.
See RoundViaDummy for more details.
References
Langsrud, Ø. and Heldal, J. (2018): “An Algorithm for Small Count Rounding of Tabular Data”. Presented at: Privacy in statistical databases, Valencia, Spain. September 26-28, 2018. https://www.researchgate.net/publication/327768398_An_Algorithm_for_Small_Count_Rounding_of_Tabular_Data
Examples
# Small example data set
z <- SmallCountData("e6")
print(z)
#> geo eu year freq
#> 1 Iceland nonEU 2018 2
#> 2 Portugal EU 2018 3
#> 3 Spain EU 2018 7
#> 4 Iceland nonEU 2019 1
#> 5 Portugal EU 2019 5
#> 6 Spain EU 2019 6
# Publishable cells by formula interface
a <- PLSrounding(z, "freq", roundBase = 5, formula = ~geo + eu + year)
print(a)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 3 0.938 1.25 1.6583
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-4 inn.5 inn.6+ inn.all pub.0 pub.1-4 pub.5 pub.6+ pub.all
#> original . 3 1 2 6 . 2 . 6 8
#> rounded 1 1 2 2 6 . . 2 6 8
#> absDiff 4 2 . . 6 3 5 . . 8
#>
print(a$inner)
#> geo year original rounded difference
#> 1 Iceland 2018 2 5 3
#> 2 Portugal 2018 3 3 0
#> 3 Spain 2018 7 7 0
#> 4 Iceland 2019 1 0 -1
#> 5 Portugal 2019 5 5 0
#> 6 Spain 2019 6 6 0
print(a$publish)
#> geo year original rounded difference
#> 1 Total Total 24 26 2
#> 2 Iceland Total 3 5 2
#> 3 Portugal Total 8 8 0
#> 4 Spain Total 13 13 0
#> 5 EU Total 21 21 0
#> 6 nonEU Total 3 5 2
#> 7 Total 2018 12 15 3
#> 8 Total 2019 12 11 -1
print(a$metrics)
#> roundBase maxRound maxdiff
#> 5.0000000 4.0000000 3.0000000
#> inner_HDutility HDutility inner_meanAbsDiff
#> 0.8131709 0.9380433 0.6666667
#> meanAbsDiff inner_rootMeanSquare rootMeanSquare
#> 1.2500000 1.2909944 1.6583124
print(a$freqTable)
#> inn.0 inn.1-4 inn.5 inn.6+ inn.all pub.0 pub.1-4 pub.5 pub.6+ pub.all
#> original 0 3 1 2 6 0 2 0 6 8
#> rounded 1 1 2 2 6 0 0 2 6 8
#> absDiff 4 2 0 0 6 3 5 0 0 8
# Using FormulaSelection()
FormulaSelection(a$publish, ~eu + year)
#> geo year original rounded difference
#> 1 Total Total 24 26 2
#> 5 EU Total 21 21 0
#> 6 nonEU Total 3 5 2
#> 7 Total 2018 12 15 3
#> 8 Total 2019 12 11 -1
FormulaSelection(a, ~eu + year) # same as above
#> geo year original rounded difference
#> 1 Total Total 24 26 2
#> 5 EU Total 21 21 0
#> 6 nonEU Total 3 5 2
#> 7 Total 2018 12 15 3
#> 8 Total 2019 12 11 -1
FormulaSelection(a) # just a$publish
#> geo year original rounded difference
#> 1 Total Total 24 26 2
#> 2 Iceland Total 3 5 2
#> 3 Portugal Total 8 8 0
#> 4 Spain Total 13 13 0
#> 5 EU Total 21 21 0
#> 6 nonEU Total 3 5 2
#> 7 Total 2018 12 15 3
#> 8 Total 2019 12 11 -1
# Recalculation of maxdiff, HDutility, meanAbsDiff and rootMeanSquare
max(abs(a$publish[, "difference"]))
#> [1] 3
HDutility(a$publish[, "original"], a$publish[, "rounded"])
#> [1] 0.9380433
mean(abs(a$publish[, "difference"]))
#> [1] 1.25
sqrt(mean((a$publish[, "difference"])^2))
#> [1] 1.658312
# Five lines below produce equivalent results
# Ordering of rows can be different
PLSrounding(z, "freq", dimVar = c("geo", "eu", "year"))
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
PLSrounding(z, "freq", formula = ~eu * year + geo * year)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
PLSrounding(z[, -2], "freq", hierarchies = SmallCountData("eHrc"))
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
PLSrounding(z[, -2], "freq", hierarchies = SmallCountData("eDimList"))
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
PLSrounding(z[, -2], "freq", hierarchies = SmallCountData("eDimList"), formula = ~geo * year)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
# Define publishable cells differently by making use of formula interface
PLSrounding(z, "freq", formula = ~eu * year + geo)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.931 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 2 2 8 12
#> rounded 1 . 2 3 6 1 . 3 8 12
#> absDiff 4 2 . . 6 8 4 . . 12
#>
# Define publishable cells differently by making use of hierarchy interface
eHrc2 <- list(geo = c("EU", "@Portugal", "@Spain", "Iceland"), year = c("2018", "2019"))
PLSrounding(z, "freq", hierarchies = eHrc2)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9357 0.2667 0.5164
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 2 2 11 15
#> rounded 1 . 2 3 6 1 . 3 11 15
#> absDiff 4 2 . . 6 11 4 . . 15
#>
# Also possible to combine hierarchies and formula
PLSrounding(z, "freq", hierarchies = SmallCountData("eDimList"), formula = ~geo + year)
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 0 1 0 0
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . . 2 7 9
#> rounded . 2 1 3 6 . . 2 7 9
#> absDiff 6 . . . 6 9 . . . 9
#>
# Single data frame output
PLSroundingInner(z, "freq", roundBase = 5, formula = ~geo + eu + year)
#> geo year original rounded difference
#> 1 Iceland 2018 2 5 3
#> 2 Portugal 2018 3 3 0
#> 3 Spain 2018 7 7 0
#> 4 Iceland 2019 1 0 -1
#> 5 Portugal 2019 5 5 0
#> 6 Spain 2019 6 6 0
PLSroundingPublish(z, roundBase = 5, formula = ~geo + eu + year)
#> geo year original rounded difference
#> 1 Total Total 6 5 -1
#> 2 Iceland Total 2 0 -2
#> 3 Portugal Total 2 0 -2
#> 4 Spain Total 2 5 3
#> 5 EU Total 4 5 1
#> 6 nonEU Total 2 0 -2
#> 7 Total 2018 3 5 2
#> 8 Total 2019 3 0 -3
# Microdata input
PLSroundingInner(rbind(z, z), roundBase = 5, formula = ~geo + eu + year)
#> geo year original rounded difference
#> 1 Portugal 2018 2 0 -2
#> 2 Spain 2018 2 5 3
#> 3 Iceland 2018 2 0 -2
#> 4 Portugal 2019 2 0 -2
#> 5 Spain 2019 2 0 -2
#> 6 Iceland 2019 2 5 3
# Zero perturbed due to both extend0 = TRUE and zeroCandidates = TRUE
set.seed(12345)
PLSroundingInner(z[sample.int(5, 12, replace = TRUE), 1:3],
formula = ~geo + eu + year, roundBase = 5,
extend0 = TRUE, zeroCandidates = TRUE, printInc = TRUE)
#> [preAggregate 12*3->5*4]
#> [extend0 5*4->6*4]
#> [-**..:=]
#> geo year original rounded difference
#> 1 Portugal 2018 4 0 -4
#> 2 Spain 2018 3 0 -3
#> 3 Iceland 2018 2 5 3
#> 4 Portugal 2019 1 0 -1
#> 5 Iceland 2019 2 0 -2
#> 6 Spain 2019 0 5 5
# Parameter avoidHierarchical (see RoundViaDummy and ModelMatrix)
PLSroundingPublish(z, roundBase = 5, formula = ~geo + eu + year, avoidHierarchical = TRUE)
#> geo eu year original rounded difference
#> 1 Total Total Total 6 5 -1
#> 2 Iceland Total Total 2 0 -2
#> 3 Portugal Total Total 2 0 -2
#> 4 Spain Total Total 2 5 3
#> 5 Total EU Total 4 5 1
#> 6 Total nonEU Total 2 0 -2
#> 7 Total Total 2018 3 5 2
#> 8 Total Total 2019 3 0 -3
# To illustrate hierarchical_extend0
# (parameter to underlying function, SSBtools::Extend0fromModelMatrixInput)
PLSroundingInner(z[-c(2:3), ], roundBase = 5, formula = ~geo + eu + year,
avoidHierarchical = TRUE, zeroCandidates = TRUE, extend0 = TRUE)
#> geo eu year original rounded difference
#> 1 Iceland nonEU 2018 1 0 -1
#> 2 Portugal EU 2019 1 0 -1
#> 3 Spain EU 2019 1 0 -1
#> 4 Iceland nonEU 2019 1 0 -1
#> 5 Portugal nonEU 2018 0 0 0
#> 6 Spain nonEU 2018 0 0 0
#> 7 Iceland EU 2018 0 0 0
#> 8 Portugal EU 2018 0 0 0
#> 9 Spain EU 2018 0 0 0
#> 10 Portugal nonEU 2019 0 0 0
#> 11 Spain nonEU 2019 0 0 0
#> 12 Iceland EU 2019 0 5 5
PLSroundingInner(z[-c(2:3), ], roundBase = 5, formula = ~geo + eu + year,
avoidHierarchical = TRUE, zeroCandidates = TRUE, extend0 = TRUE,
hierarchical_extend0 = TRUE)
#> geo eu year original rounded difference
#> 1 Iceland nonEU 2018 1 0 -1
#> 2 Portugal EU 2019 1 0 -1
#> 3 Spain EU 2019 1 0 -1
#> 4 Iceland nonEU 2019 1 5 4
#> 5 Portugal EU 2018 0 0 0
#> 6 Spain EU 2018 0 0 0
# Package sdcHierarchies can be used to create hierarchies.
# The small example code below works if this package is available.
if (require(sdcHierarchies)) {
z2 <- cbind(geo = c("11", "21", "22"), z[, 3:4], stringsAsFactors = FALSE)
h2 <- list(
geo = hier_compute(inp = unique(z2$geo), dim_spec = c(1, 1), root = "Tot", as = "df"),
year = hier_convert(hier_create(root = "Total", nodes = c("2018", "2019")), as = "df"))
PLSrounding(z2, "freq", hierarchies = h2)
}
#> Loading required package: sdcHierarchies
#> Loading required package: shinythemes
#> Package 'sdcHierarchies' 0.21.0 has been loaded.
#>
#> PLSrounding summary:
#>
#> maxdiff HDutility meanAbsDiff rootMeanSquare
#> 1 0.9117 0.3333 0.5774
#>
#> Frequencies of cell frequencies and absolute differences:
#>
#> inn.0 inn.1-2 inn.3 inn.4+ inn.all pub.0 pub.1-2 pub.3 pub.4+ pub.all
#> original . 2 1 3 6 . 4 3 11 18
#> rounded 1 . 2 3 6 2 . 5 11 18
#> absDiff 4 2 . . 6 12 6 . . 18
#>
# Use PLS2way to produce tables as in Langsrud and Heldal (2018) and to demonstrate
# parameters maxRound, zeroCandidates and identifyNew (see RoundViaDummy).
# Parameter rndSeed used to ensure same output as in reference.
exPSD <- SmallCountData("exPSD")
a <- PLSrounding(exPSD, "freq", 5, formula = ~rows + cols, rndSeed=124)
PLS2way(a, "original") # Table 1
#> col1 col2 col3 col4 col5 Total
#> row1 6 0 1 3 4 14
#> row2 1 2 3 1 2 9
#> row3 0 1 1 0 2 4
#> Total 7 3 5 4 8 27
PLS2way(a) # Table 2
#> col1 col2 col3 col4 col5 Total
#> row1 6 0 5 0 4 15
#> row2 1 0 0 5 2 8
#> row3 0 5 0 0 0 5
#> Total 7 5 5 5 6 28
a <- PLSrounding(exPSD, "freq", 5, formula = ~rows + cols, identifyNew = FALSE, rndSeed=124)
PLS2way(a) # Table 3
#> col1 col2 col3 col4 col5 Total
#> row1 6 0 1 0 4 11
#> row2 1 0 3 5 2 11
#> row3 0 5 0 0 0 5
#> Total 7 5 4 5 6 27
a <- PLSrounding(exPSD, "freq", 5, formula = ~rows + cols, maxRound = 7)
PLS2way(a) # Values in col1 rounded
#> col1 col2 col3 col4 col5 Total
#> row1 5 0 0 5 0 10
#> row2 0 0 5 0 5 10
#> row3 0 5 0 0 0 5
#> Total 5 5 5 5 5 25
a <- PLSrounding(exPSD, "freq", 5, formula = ~rows + cols, zeroCandidates = TRUE)
PLS2way(a) # (row3, col4): original is 0 and rounded is 5
#> col1 col2 col3 col4 col5 Total
#> row1 6 0 5 0 4 15
#> row2 1 5 0 0 2 8
#> row3 0 0 0 5 0 5
#> Total 7 5 5 5 6 28
# Using formula followed by FormulaSelection
output <- PLSrounding(data = SmallCountData("example1"),
formula = ~age * geo * year + eu * year,
freqVar = "freq",
roundBase = 5)
FormulaSelection(output, ~(age + eu) * year)
#> age geo year original rounded difference
#> 1 Total Total Total 59 59 0
#> 2 old Total Total 38 37 -1
#> 3 young Total Total 21 22 1
#> 7 Total Total 2014 20 21 1
#> 8 Total Total 2015 18 16 -2
#> 9 Total Total 2016 21 22 1
#> 10 Total EU Total 46 49 3
#> 11 Total nonEU Total 13 10 -3
#> 18 old Total 2014 13 16 3
#> 19 old Total 2015 13 11 -2
#> 20 old Total 2016 12 10 -2
#> 21 young Total 2014 7 5 -2
#> 22 young Total 2015 5 5 0
#> 23 young Total 2016 9 12 3
#> 33 Total EU 2014 15 16 1
#> 34 Total nonEU 2014 5 5 0
#> 35 Total EU 2015 15 16 1
#> 36 Total nonEU 2015 3 0 -3
#> 37 Total EU 2016 16 17 1
#> 38 Total nonEU 2016 5 5 0
# Example similar to the one in the documentation of tables_by_formulas,
# but using PLSroundingPublish with roundBase = 4.
tables_by_formulas(SSBtoolsData("magnitude1"),
table_fun = PLSroundingPublish,
table_formulas = list(table_1 = ~region * sector2,
table_2 = ~region1:sector4 - 1,
table_3 = ~region + sector4 - 1),
substitute_vars = list(region = c("geo", "eu"), region1 = "eu"),
collapse_vars = list(sector = c("sector2", "sector4")),
roundBase = 4)
#> region sector original rounded difference table_1 table_2 table_3
#> 1 Total Total 20 21 1 TRUE FALSE FALSE
#> 2 Iceland Total 4 4 0 TRUE FALSE TRUE
#> 3 Portugal Total 8 8 0 TRUE FALSE TRUE
#> 4 Spain Total 8 9 1 TRUE FALSE TRUE
#> 5 EU Total 16 17 1 TRUE FALSE TRUE
#> 6 nonEU Total 4 4 0 TRUE FALSE TRUE
#> 7 Total private 16 17 1 TRUE FALSE FALSE
#> 8 Total public 4 4 0 TRUE FALSE FALSE
#> 9 Total Agriculture 4 4 0 FALSE FALSE TRUE
#> 10 Total Entertainment 6 5 -1 FALSE FALSE TRUE
#> 11 Total Governmental 4 4 0 FALSE FALSE TRUE
#> 12 Total Industry 6 8 2 FALSE FALSE TRUE
#> 13 Iceland private 4 4 0 TRUE FALSE FALSE
#> 14 Portugal private 6 4 -2 TRUE FALSE FALSE
#> 15 Portugal public 2 4 2 TRUE FALSE FALSE
#> 16 Spain private 6 9 3 TRUE FALSE FALSE
#> 17 Spain public 2 0 -2 TRUE FALSE FALSE
#> 18 EU private 12 13 1 TRUE FALSE FALSE
#> 19 EU public 4 4 0 TRUE FALSE FALSE
#> 20 nonEU private 4 4 0 TRUE FALSE FALSE
#> 21 EU Agriculture 4 4 0 FALSE TRUE FALSE
#> 22 EU Entertainment 5 5 0 FALSE TRUE FALSE
#> 23 EU Governmental 4 4 0 FALSE TRUE FALSE
#> 24 EU Industry 3 4 1 FALSE TRUE FALSE
#> 25 nonEU Entertainment 1 0 -1 FALSE TRUE FALSE
#> 26 nonEU Industry 3 4 1 FALSE TRUE FALSE