Closed tompollard closed 6 years ago
Tukey's rule is pretty common, although the interpretation depends on the circumstances.
I may just apply this rule, and suggest the user plot the data to check.
Thanks Jesse. Tukey's rule will be straightforward to incorporate, assuming you mean this IQR based approach: https://en.wikipedia.org/wiki/Outlier#Tukey's_fences
Pull request: https://github.com/tompollard/tableone/pull/44 adds Tukey outlier test to the cont_describe
table:
Also raises warnings if outliers are detected:
Related to #19, #22 and #28. Might also be worth raising a warning if outliers appear to be heavily influencing the summary statistics. What is a good approach? (e.g. simple check might be to compare trimmed mean with mean?).