Create new variables using the median

[Ihab Man]

Dear all.
Please, How can I create new two variables (Z and Y) using the median of variable (X) ? . I mean I want to create variable Z equal to the values above the median of variable X and Y otherwise. Thanks a lot.

[Nick Cox]

If you

summarize X, detail

then the median of X is accessible afterwards as r(p50) to refer to in later commands. Sometimes it is worth putting this in a scalar for safety say

scalar Xmedian = r(p50)

I don't understand your definition confidently but you seem to be asking for

gen Z = cond(X > Xmedian, X, Y)

[Sergiy Radyakin]

clear all
sysuse auto

centile price
local median=r(c_1)
generate expensive=price if price>=`median'
generate cheap=price if price <`median'

sort price
list price cheap expensive

[Sergiy Radyakin]

Crossed with Nick's answer, but I understood the question differently, in that it should be TWO new variables from ONE source, not ONE new variable from TWO sources. The question was not very precisely written.

[Nick Cox]

Indeed, "two new variables" could mean #3 although in that case separate would work too.

[Ihab Man]

Dear Nick Cox & Sergiy Radyakin
Thanks for your help.
I mean I have one variable called X and I want to create two variables form X using the median of it. One of them equal to and above the median of it and other one will be the values below the median of it . Thanks

[Nick Cox]

OK, so either that is already answered by #3 or #5 or you are still unclear. Which is it?

[Ihab Man]

Dear Nick Cox
No . Every thing is clear now . I thought that you are still not understand my question. Now OK. Thanks a lot and of Course Thanks a lot Dear Sergiy Radyakin.

[Nick Cox]

Indeed; it's a parsing question: where do the braces { } go?

I mean I want to create variable Z equal to { the values above the median of variable X and Y otherwise}.

I mean I want to create variable Z equal to the {values above the median of variable X} and Y otherwise.

My contribution to the question was writing

separate

....

[Ihab Man]

Dear Nick Cox . I am Sorry . YES, it was my mistake. I was meaning for (Y otherwise ) that Y equal to the values below the median of X.

[Nick Cox]

All fine now.

[Ihab Man]

Dear Nick Cox
Please, I understood the comment from author wrong . Please, I have panel data ( 500 firms , 10 years) . I want to divide the firms into two group based on one of my independent variable "X1" ( using the median of X1) . Group 1 = banks below the median of X1 and group 2 = banks above the median of X1. Please, how can I do it using STATA? Thanks a lot.

[Nick Cox]

#3 would be my answer to #12, which I don't understand to be a different question.