Creating Dummy variables for all Variables on the basis of median values

[Attaullah Shah]

I have 10 variables and want to create dummy for them based on their median values, is there any easy solution or loop to do that.
Thanks
Attaullah Shah

[Mike Murphy]

It depends what you mean by

based on their median values

, if you wanted an indicator for, say, whether they were less than the median, you could do something like:

local varnames = "var1 var2 ... var10"

foreach var of local varnames {

egen temp = median(`var')
gen `var'_dummy = 0
replace `var'_dummy = 1 if `var' < temp
drop temp

}

Mike

[Aspen Chen]

An example:

sysuse auto, clear
foreach var of varlist rep78 turn    {
    qui su `var',d
    gen `var'_m=`var'==`r(p50)'
}

[Nick Cox]

Aspen Chen: I think you want >= not ==

[Aspen Chen]

Nick, I really thought the OP had meant "exactly the median". But you're probably right. The >= qualifier seems a more useful application.

[Attaullah Shah]

Thank you Aspen Chen, mmurphy, and Nick for your wonderful solution. One last question, how would the code treat missing values, do we need to make some adjustment in that code. My best guess would be, for example in the Aspen Chen code, to use

sysuse auto, clear
foreach var of varlist rep78 turn {
qui su `var',d
gen `var'_m=`var'>=`r(p50)' if `var'!=.
}

[Clyde Schechter]

Or perhaps simpler is

foreach v of varlist [whatever] {
xtile `v'_msplit = `v', nq(2)
}

These indicator variables will be coded as 1 and 2 rather than 0/1, but you can always subtract 1 if that's necessary.