RoundKostra: Rounding à la Heldal following the standards in the Kostra...
In statisticsnorway/Kostra: Functions for Kostra
RoundKostra R Documentation
Rounding à la Heldal following the standards in the Kostra project.
Description
Rounding à la Heldal following the standards in the Kostra project.
Usage
RoundKostra(
data,
idVar,
strataVar = NULL,
freqVar,
freqVarGroup = NULL,
roundBase = 3,
method = "pls",
formula = NULL,
level = 0,
allSmall = TRUE,
singleTotal = TRUE,
makeSums = TRUE,
output = "rounded",
total = "Total",
split = "_",
extraOutput = FALSE,
seed = 12345,
...
)
Arguments
data
Input data set of class data.frame
idVar
Id-variable (name or number)
strataVar
Strata-variable(s) (name or number)
freqVar
Variable(s) holding counts (name or number)
freqVarGroup
NULL (default) or integer representing groups of variables (see details)
roundBase
Basis for rounding
method
Algorithm for the rounding calculations. Currently "pls" or "singleRandom".
formula
Model formula as a string defining cells to be published (additional to automation)
level
Interaction level or 0 (all levels) defining complexity of model component created from strata
allSmall
When TRUE all small values will be rounded (when a single freqVar)
singleTotal
When TRUE identical rowsums in all freqVarGroups needed.
When FALSE totals for each freqVarGroup will be in output.
makeSums
When TRUE totals vil be made similar to ProtectKostra (in fact this is done by calling ProtectKostra)
output
One of "rounded" (default), "original", "difference" or "status".
Unrounded results are returned by “original” (same as roundBase=0) and
"difference" = "rounded" - “original”.
With output=”status” zero differences are set to “o” (original) and others are coded as “r” (rounded)
total
String used to name totals.
split
Parameter to AutoSplit - see varNames and rowData above.
When NULL automatic splitting without needing a split string.
extraOutput
When TRUE output is a list of several elements (makeSums ignored)
seed
NULL or seed for random number generator (set.seed(seed) will be run at the beginning of the function)
...
Variables for formula and additional variables that will be included in output (name or number).
Details
A single freq variable and formula:
A formula defines all the publishable cells. Rounding is performed so that all the publishable cells are safe.
When allSmall=TRUE all small cells are rounded. All possible totals are then safe, but totals not defined by
the formula can be far from the original values.
A single freq variable and strata:
Instead of a formula it is assumed that the cells to be published are obtained by crossing all strata variables.
The parameter "level" may be used.
Several freq variables without freqVarGroup:
The data is in a unstacked form and stack/unstack will be performed in the background similar to ProtectKostra.
The original id-var will be considered as a strata-var when stacked and rounding is performed similar
to "A single freq variable and strata".
With freqVarGroup without single-groups:
Each group can be stacked to form a separate data set, but a common data set is needed.
An ad hoc data set is created to match all the single data sets and this data set will be used in the rounding process.
This method will not work in complicated cases. Use with care.
Try extraOutput=TRUE to see what is going on.
With freqVarGroup with single-groups:
The single-groups are assumed to be two-category groups (yes and no, but only yes is reported).
The remaining category will be computed in order to create ad hoc data set.
Value
A data.frame unless extraOutput = TRUE
Note
Even if freq-variables with freqVarGroup<1 is not used they will be read by GetData together with the other
the freq-variables variables into a matrix. Use a common numeric type for all these variables to prevent change of data type.
Parameter namesAsInput in ProtectTable() is not yet available in RoundKostra and therefore very advanced variable name
coding will give insufficient results. Example:
RoundKostra(KostraData("z3wb") ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15)
But this will work:
RoundKostra(KostraData("z3wb") ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,split=NULL)
See similar example in ProtectTable.
NOTE: Be sure to spell the input parameters correctly. Because of the "..."-input misspelled parameters
may give strange results instead of error.
See Also
ProtectKostra, RoundViaDummy, makeroundtabs,
Round2, FormulaSums, ModelMatrix
Examples
# ========================================================
# Examples: A single freq variable
# =======================================================
z2w <- KostraData("z2w")
# ==== Without strataVar and without formula ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", roundBase=5)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", makeSums=FALSE) # Without total
# ==== With strataVar and without formula ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), roundBase=5)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), makeSums=FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar= "fylke", allSmall=FALSE) # Warning when makeSums=TRUE
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar= "fylke", allSmall=FALSE, makeSums=FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), extraOutput=TRUE)$formula
# ==== Without strataVar and with formula ( makeSums ignored without warning) ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="fylke", fylke="fylke")
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="fylke", fylke="fylke",allSmall = FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="A+B", A="fylke",B="kostragr")
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="A*B+C", A="fylke",B="kostragr",C="annet")
# =============================================================================
# Examples: Several freq variables without freqVarGroup (allSmall ignored)
# =============================================================================
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7)
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,makeSums=FALSE)
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,extraOutput=TRUE)$input
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,extraOutput=TRUE)$formula
# ===============================================================
# Examples: With freqVarGroup
# ==========================================================================
# ========== With no single-groups ================
ex1 = Kostra:::exData1() # hack endre seinere
freqVarGroup <- c(1,1,1,1,1,1,1,1,2,2,2,2)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup, makeSums=FALSE)
a1 <- RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup, extraOutput=TRUE)
head(a1$input) # ad hoc created data
a1$formula # The formula used
# ========== With some single-groups ================
freqVarGroup <- c(1,1,1,1,1,1,1,1,2,2,2,2,3,4,-1,5)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE)
# ====== With incorrect totals
ex1b <- ex1
ex1b$s1[1]=2L
ex1b$arb_A[2]=5L
RoundKostra(ex1b, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE)
a2 <- RoundKostra(ex1b, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE,extraOutput=TRUE)
table(a2$input$s1_s2_s3_s4, useNA ="always") # Missing values in ad hoc created data when incorrect totals
# ===============================================================
# Examples With parameter output
# ===============================================================
RoundKostra(z2w ,idVar="region", freqVar="arbeid",output="status")
RoundKostra(z2w ,idVar="region", freqVar="arbeid",output="difference")
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,output="status")
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,output="difference")
# ===============================================================
# Micro data example (":::" since functions not exported yet)
# ===============================================================
microData <- Kostra:::microEx1() # A micro data set
freqData <- MakeFreq(microData,"freq") # Make cross-classified data
freqData$id <- 1:NROW(freqData) # Add id-variable
# Rounding with makeSums=FALSE
freqRound <- RoundKostra(freqData, idVar="id", strataVar=c("region", "fylke", "kostragr", "hovedint"), freqVar="freq", makeSums=FALSE)
microRound <- MakeMicro(freqRound,"freq") # Create micro data set from output
microRound <- microRound[,-c(1,6)] # Remove some variables
# Alternative where only region sums and the cross-classifications fylke*hovedint and kostragr*hovedint are to be published
freqRound2 <- RoundKostra(freqData, idVar="id", formula=("region +fylke*hovedint + kostragr*hovedint"), freqVar="freq",
makeSums=FALSE, allSmall=FALSE, region="region",fylke="fylke",kostragr="kostragr",hovedint="hovedint")
microRound2 <- MakeMicro(freqRound2,"freq")
microRound2 <- microRound2[,-c(1,2)]
statisticsnorway/Kostra documentation built on Sept. 25, 2024, 10:37 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
| RoundKostra | R Documentation |
Rounding à la Heldal following the standards in the Kostra project.
Description
Rounding à la Heldal following the standards in the Kostra project.
Usage
RoundKostra(
data,
idVar,
strataVar = NULL,
freqVar,
freqVarGroup = NULL,
roundBase = 3,
method = "pls",
formula = NULL,
level = 0,
allSmall = TRUE,
singleTotal = TRUE,
makeSums = TRUE,
output = "rounded",
total = "Total",
split = "_",
extraOutput = FALSE,
seed = 12345,
...
)
Arguments
data |
Input data set of class data.frame |
idVar |
Id-variable (name or number) |
strataVar |
Strata-variable(s) (name or number) |
freqVar |
Variable(s) holding counts (name or number) |
freqVarGroup |
NULL (default) or integer representing groups of variables (see details) |
roundBase |
Basis for rounding |
method |
Algorithm for the rounding calculations. Currently "pls" or "singleRandom". |
formula |
Model formula as a string defining cells to be published (additional to automation) |
level |
Interaction level or 0 (all levels) defining complexity of model component created from strata |
allSmall |
When TRUE all small values will be rounded (when a single freqVar) |
singleTotal |
When TRUE identical rowsums in all freqVarGroups needed. When FALSE totals for each freqVarGroup will be in output. |
makeSums |
When TRUE totals vil be made similar to ProtectKostra (in fact this is done by calling ProtectKostra) |
output |
One of "rounded" (default), "original", "difference" or "status".
Unrounded results are returned by “original” (same as roundBase=0) and
|
total |
String used to name totals. |
split |
Parameter to |
extraOutput |
When TRUE output is a list of several elements (makeSums ignored) |
seed |
NULL or seed for random number generator ( |
... |
Variables for formula and additional variables that will be included in output (name or number). |
Details
A single freq variable and formula: A formula defines all the publishable cells. Rounding is performed so that all the publishable cells are safe. When allSmall=TRUE all small cells are rounded. All possible totals are then safe, but totals not defined by the formula can be far from the original values.
A single freq variable and strata: Instead of a formula it is assumed that the cells to be published are obtained by crossing all strata variables. The parameter "level" may be used.
Several freq variables without freqVarGroup: The data is in a unstacked form and stack/unstack will be performed in the background similar to ProtectKostra. The original id-var will be considered as a strata-var when stacked and rounding is performed similar to "A single freq variable and strata".
With freqVarGroup without single-groups: Each group can be stacked to form a separate data set, but a common data set is needed. An ad hoc data set is created to match all the single data sets and this data set will be used in the rounding process. This method will not work in complicated cases. Use with care. Try extraOutput=TRUE to see what is going on.
With freqVarGroup with single-groups: The single-groups are assumed to be two-category groups (yes and no, but only yes is reported). The remaining category will be computed in order to create ad hoc data set.
Value
A data.frame unless extraOutput = TRUE
Note
Even if freq-variables with freqVarGroup<1 is not used they will be read by GetData together with the other
the freq-variables variables into a matrix. Use a common numeric type for all these variables to prevent change of data type.
Parameter namesAsInput in ProtectTable() is not yet available in RoundKostra and therefore very advanced variable name
coding will give insufficient results. Example:
RoundKostra(KostraData("z3wb") ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15)
But this will work:
RoundKostra(KostraData("z3wb") ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,split=NULL)
See similar example in ProtectTable.
NOTE: Be sure to spell the input parameters correctly. Because of the "..."-input misspelled parameters may give strange results instead of error.
See Also
ProtectKostra, RoundViaDummy, makeroundtabs,
Round2, FormulaSums, ModelMatrix
Examples
# ========================================================
# Examples: A single freq variable
# =======================================================
z2w <- KostraData("z2w")
# ==== Without strataVar and without formula ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", roundBase=5)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", makeSums=FALSE) # Without total
# ==== With strataVar and without formula ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), roundBase=5)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), makeSums=FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar= "fylke", allSmall=FALSE) # Warning when makeSums=TRUE
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar= "fylke", allSmall=FALSE, makeSums=FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", strataVar=c("fylke","kostragr"), extraOutput=TRUE)$formula
# ==== Without strataVar and with formula ( makeSums ignored without warning) ====
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="fylke", fylke="fylke")
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="fylke", fylke="fylke",allSmall = FALSE)
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="A+B", A="fylke",B="kostragr")
RoundKostra(z2w ,idVar="region", freqVar="arbeid", formula="A*B+C", A="fylke",B="kostragr",C="annet")
# =============================================================================
# Examples: Several freq variables without freqVarGroup (allSmall ignored)
# =============================================================================
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7)
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,makeSums=FALSE)
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,extraOutput=TRUE)$input
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,extraOutput=TRUE)$formula
# ===============================================================
# Examples: With freqVarGroup
# ==========================================================================
# ========== With no single-groups ================
ex1 = Kostra:::exData1() # hack endre seinere
freqVarGroup <- c(1,1,1,1,1,1,1,1,2,2,2,2)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup, makeSums=FALSE)
a1 <- RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:15,freqVarGroup=freqVarGroup, extraOutput=TRUE)
head(a1$input) # ad hoc created data
a1$formula # The formula used
# ========== With some single-groups ================
freqVarGroup <- c(1,1,1,1,1,1,1,1,2,2,2,2,3,4,-1,5)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup)
RoundKostra(ex1, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE)
# ====== With incorrect totals
ex1b <- ex1
ex1b$s1[1]=2L
ex1b$arb_A[2]=5L
RoundKostra(ex1b, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE)
a2 <- RoundKostra(ex1b, idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:19,freqVarGroup=freqVarGroup, singleTotal = FALSE,extraOutput=TRUE)
table(a2$input$s1_s2_s3_s4, useNA ="always") # Missing values in ad hoc created data when incorrect totals
# ===============================================================
# Examples With parameter output
# ===============================================================
RoundKostra(z2w ,idVar="region", freqVar="arbeid",output="status")
RoundKostra(z2w ,idVar="region", freqVar="arbeid",output="difference")
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,output="status")
RoundKostra(z2w ,idVar="region",strataVar=c("fylke","kostragr"),freqVar=4:7,output="difference")
# ===============================================================
# Micro data example (":::" since functions not exported yet)
# ===============================================================
microData <- Kostra:::microEx1() # A micro data set
freqData <- MakeFreq(microData,"freq") # Make cross-classified data
freqData$id <- 1:NROW(freqData) # Add id-variable
# Rounding with makeSums=FALSE
freqRound <- RoundKostra(freqData, idVar="id", strataVar=c("region", "fylke", "kostragr", "hovedint"), freqVar="freq", makeSums=FALSE)
microRound <- MakeMicro(freqRound,"freq") # Create micro data set from output
microRound <- microRound[,-c(1,6)] # Remove some variables
# Alternative where only region sums and the cross-classifications fylke*hovedint and kostragr*hovedint are to be published
freqRound2 <- RoundKostra(freqData, idVar="id", formula=("region +fylke*hovedint + kostragr*hovedint"), freqVar="freq",
makeSums=FALSE, allSmall=FALSE, region="region",fylke="fylke",kostragr="kostragr",hovedint="hovedint")
microRound2 <- MakeMicro(freqRound2,"freq")
microRound2 <- microRound2[,-c(1,2)]
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.