Checks that conf.level is a single number in \((0, 1)\), or
NA. Intended for the confidence-interval functions across the
suite, so that all of them accept the same values and refuse the rest
with the same message.
Value
conf.level, invisibly,
so the check can be used in an
assignment: conf.level <- checkConfLevel(conf.level).
Details
The order of the tests is the point of this function. NA is
logical, so a check that leads with !is.numeric() rejects
the very default most of these functions carry. And is.na() on a
vector of length other than one turns the surrounding if into the
error message, which then talks about the condition instead of the
argument. Length first, then type, then range.
NaN is excluded explicitly: is.na(NaN) is TRUE, so
without that test a NaN would be silently accepted as "no
interval wanted".
Examples
checkConfLevel(0.95)
checkConfLevel(NA)
if (FALSE) { # \dontrun{
checkConfLevel(c(0.9, 0.95)) # length
checkConfLevel(NULL) # length
checkConfLevel(NaN) # not a level, and not NA either
checkConfLevel(0) # range is open
} # }