Turns a relative standard error (RSE, in percent) into a plain-language reliability rating, using a five-band scheme.
Arguments
- rse
Relative standard error(s), in percent (see
rse()).
Details
The bands are: under 5% "high precision", under 10% "precise", under 15%
"satisfactory", under 25% "use with caution", and 25% or more "likely
reliability issues". Rating estimates by their relative standard error is how
national statistical agencies signal whether a survey number is solid enough
to publish: the Australian Bureau of Statistics treats an RSE of 25% or more
as "not reliable for most purposes", flags 25–50% with an asterisk as
estimates to use with caution, and suppresses anything above 50%. The 25%
boundary here is that same one. The tighter bands below it are specific to
consumer-survey work, where the question is usually which of several
adequately precise numbers to lead a report with. This drives the precision
column of diagnose() and the overall verdict of precision_summary().
See also
rse(), diagnose(), precision_summary().
Other diagnostics:
diagnose(),
margin_of_error(),
precision_summary(),
rse(),
se_mean(),
se_prop()
Examples
rse_rating(c(3, 8, 12, 20, 40))
#> [1] "high precision" "precise"
#> [3] "satisfactory" "use with caution"
#> [5] "likely reliability issues"