Check, and attempt to adjust, synthetic datasets NOT created by syn()
if not compatible with the original.
The output is a list with 4 components, the first two giving adjusted
versions of the input synthetic data and original and the third needsfix
an indicator of whether the functions listed below are likely to
run correctly on the adjusted data, the fourth unchanged indicates
whether any changes have been made to the original or the synthetic data.
Variables that are in both the synthetic and the original data are checked to 1) convert any character variables to R factors 2) check that data types match 3) check differences in whether variables have missing values 4) check if the levels of factors agree and if not use the combination of both sets.
needsfix becomes TRUE if 1) some variables are in the synthetic but
not in the original 2) variables have different classes (after characters
converted to factor) 3) there are missing values in the synthetic data
but not in the original 4) some levels of a factor are in the synthetic data
but not in the original.
Some warning messages are printed if level differences are not just due to
missing values.
The function can be run to compare original and synthetic or it can be called
by setting compare.synorig = TRUE when data are not supplied as an object
of class synds created by synthpop in the following functions :
utility.tab() utility.tables() compare() disclosure() disclosure.summary().
In this case the function attempts to correct differences and continue or, if this
is impossible, will prompt the user as to which variables need changing.
synorig.compare(syn,orig, print.flag = TRUE)A list with 3 components
adjusted version of syn or of the first element of
the list syn if it is a list.
adjusted version of orig.
TRUE/FALSE as to whether the outputs need to be fixed before utility and disclosure functions could be used on them.
TRUE/FALSE indicating if the outputs of the function are unchanged from the outputs.
A data set containing the synthesised data, or a list of such data
sets. When syn is a list only the first member of the list is used and
the syn component of the result is the adjusted version of this first member.
The original data set.
If TRUE prints non-essential summary messages.
Error messages explain briefly what adjustments have been made to the data
sets, what could not be fixed and what might need to be checked.
Both orig and syn are made
into simple data frames for comparison (e.g. if tibbles or matrices)
to add
utility.gen utility.tab utility.tables
compare.synds disclosure.synds
library(synthpop)
orig <- SD2011[1:2000,]
pretendsyn <- SD2011[2001:5000, 1:5]
orig[,1] <- as.character(orig[,1])
codebook.syn(orig[,1:5])
newdata <- synorig.compare(pretendsyn, orig)
codebook.syn(newdata$orig[,1:5])
Run the code above in your browser using DataLab