Outliers and winsorization

[Amina Sagittaire]

Dear all,

I am pretty new to Stata so I am wondering when to winsorize my variables. Boxplots indicate that my DV, my IV and some controls do indeed have outliers. However, is there a more tangible way to identify outliers? Also, do I winsorize all such variables or do I focus on DV and IV only?

Thanks a lot in advance!

[Nick Cox]

There are several threads here on this technique and its disadvantages. Please search the forum.

[Carlo Lazzaro]

Dear Amina:
welcome to this forum.
Unless you're 100% sure that the so called outliers are the offspring of a mistaken data entry, deleting observations means throwing out pieces of information.
Data generating process may well include observations that live in the outskirts of the end of the right tail (and legally so).
See @Nick Cox's humorous example on baketball players at https://www.statalist.org/forums/for...smaller-than-1 #2.

PS: A bit late to the party!