Check a GADSdat for large labeled whole-number values.
Source: R/checkIntOverflow.R
checkIntOverflow.RdCheck a GADSdat object for any occurrences of labeled whole-number values
that would be too large for R to handle if they were coerced as.integer().
Value
Returns a data.frame, listing the affected varNames,
the large whole-number values, their respective missings tags,
and whether they actually occur in the data (empty).
The rownums of the affected rows in GADSdat$labels are also
provided in a separate column as a fail safe.
Details
According to its documentation, R can only handle integer
values of up to (roughly) \(\pm 2 \times 10^9\) (2,147,483,647 to be exact;
c.f. .Machine$integer.max).
This restriction is only relevant when exporting a GADSdat to .dta
and only when any value exceeding the limit is also labeled (or tagged as missing).
This is because Stata only accepts labeled integer (not labeled floating-point;
c.f. checkLabeledFractionals in this package)
values. haven's write_dta function will therefore
try to coerce any labeled values as.integer(). Unlabeled values, however, will
stay generic numeric values that have a higher limit.
Examples
# Introduce a fractional value into meta data
pisa2 <- changeMissings(GADSdat = pisa,
varName = "schtype",
value = 9999999999,
missings = "miss")
checkIntOverflow(pisa2)
#> varName value missings empty rownum
#> 1 schtype 1e+10 miss TRUE 7