Boxplot can be used to identify outliers, but if the baseline is shifting upwards, the result is biased. Thus, I am wondering how to detect outliers in such a scenario? Thanks in advance.
You can screen for +/- 2 std, etc, but there is no substitute for visual inspection. At Morgan Stanley, we looked at 25,000+ time series every time we updated the data. There may be an AI/ML solution, but a) I haven’t seen one and b) I wouldn’t want to bet my career on it.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.