Wisorizing variable at the upper and lower 1%

[Badar Khalid]

Dear Statalist Users - I would like winsorize the sales growth variables at upper and lower 1%. The variable is summarized below:

Variable	Obs	Mean	Std. Dev.	Min	Max
SGrowth	3,000	.16	2.5	-7.22	14.7067

Can you please provide me with the exact command to do so. Thank you in advance

[ben earnhart]

"help winsor" or "search winsor" should get you going. Now whether you *should* winsorize is a different issue.

[Badar Khalid]

Thank you ben for your answer. Yes I need to winsorize some variables in my data set following particular paper in which the author has wisorized these variables.

I am wondering if I can wisorize the variable from both tails as when I have searched in wisorize help, it seems that wisorizing need to be don just from one side (upper or lower) of the a particular variable.

I am wondering if I can do the command at the same time when I am winsorizing the variable at the upper and lower 1%.

Waiting for response in this regard please. Thank you again

[Nick Cox]

"Thank you in advance" and "waiting for response" won't encourage anyone to help you or help you faster than they would otherwise. On the other hand, FAQ Advice #12 is explicit that you should say exactly what commands you are using and where they come from. Making your question clear and easy to answer will encourage members much, much more.

You refer to a wisorize help; I don't believe that there is any such program. The technique is named for Charles P. Winsor. Winsorize is a correct and standard spelling. I can't find any such program, even correcting to winsorize.

I can find

winsor from http://fmwww.bc.edu/RePEc/bocode/w

winsor2 from http://fmwww.bc.edu/RePEc/bocode/w

robmean from http://www.ats.ucla.edu/stat/stata/ado/analysis

To get advice on a particular program you must be clear on what it is.

[ben earnhart]

I think the package description is pretty clear.

TITLE
'WINSOR': module to Winsorize a variable

DESCRIPTION/AUTHOR(S)

winsor takes the non-missing values of a variable and generates a
new variable identical except that the h highest and h lowest
values are replaced by the next value counting inwards from the
extremes. h can be specified directly or indirectly by
specifying a fraction p of the number of observations: h = [p
n]. This transformation is named after the biostatistician C.P.
Winsor. This is version 1.3.0 of the software.

Author: Nicholas J. Cox, University of Durham
Support: email [email protected]

Distribution-Date: 20060809

[Badar Khalid]

Thank you ben and Nick for your answers