Winsor1 vs Winsor 2

[Wessel de Kroo]

Hi all,

What is the difference between Winsor 1 and Winsor 2? in the Stata packages.

[Nick Cox]

I guess you're referring to winsor (SSC) and winsor2 (SSC). I don't know of any winsor1 (and naturally winsor 1 would not be a legal command name).

I wrote winsor in 1998 and last revised it in 2002. I am fuzzy about the context but I think the major factor was just that people asked repeatedly for winsorizing code and it didn't seem difficult to write a simple program.

I am interested in winsorizing as a way to approach univariate distributions -- and that is linked to my later, more enthusiastic and more developed work on trimming, as written up in

SJ-13-3 st0313 . . . . . . . . . . . . . . Speaking Stata: Trimming to taste
(help trimmean, trimplot if installed) . . . . . . . . . . N. J. Cox
Q3/13 SJ 13(3):640--666
tutorial review of trimmed means, emphasizing the scope for
trimming to varying degrees in describing and exploring data

I am fairly hostile to its unthinking use before regression or similar procedures. The code is indifferent to my attitudes, naturally, but I never wanted to provide a fully fledged winsorizing program

I've never communicated (that I recall) with the author of winsor2 who acknowledges using my code. That's fine by me. He or she offers a by() option and a replace option; the latter is utterly alien to my philosophy of data analysis, which includes the principle of the original data always remaining accessible. There are other handles that look as if they could be useful. But winsor2 isn't a superset of winsor and neither I nor its author regard it as superseding winsor.

[Wessel de Kroo]

Okay, yes I was referring to

winsor

(SSC) and

winsor2

.

Clear answer and I read your stance on a different thread on using winsorizing in general. This provides some nice food for thought in upcoming analyses. Thank you.