Dividing panel by quartiles

[Ritika Khurana]

Hello everyone,

I am dividing my panel data into 4 quartiles by country's GDP and year. A country may or may not be in the same quartile each year. But, for most part they will belong to same quartile. This dataset has 130 countries and 33 periods starting from 1984 to 2016.
For this, I run the following code:

 sort cty year

. forval x = 25(25) 75 {
  2. by year : egen p`x' = pctile(rgdpna), p(`x')
  3. }

rgdpna is the real GDP of a country. cty = country

However, I get an error message.

not sorted
r(5);

.

I tried including -sort- command in the -for- loop but didn't work either.

Could you tell me where am I going wrong?

[David Radwin]

Try

sort year cty

[Ritika Khurana]

David Radwin Thank you replying. I tried. But, there was another error message.

 sort  year cty

. forval x = 25(25) 75 {
  2. by year : egen p`x' = pctile(rgdpna), p(`x')
  3. }
variable p25 already defined
r(110);

[David Radwin]

You'll need to drop that variable (p25) before you generate it again. Same for other variables that may exist. Here is a shortcut:

forval x = 25(25) 75 {
     capture drop p`x'
     by year : egen p`x' = pctile(rgdpna), p(`x')
     }

[Nick Cox]

Note that using the xtile function for egen in egenmore (SSC) is another way to get quartile bins as you evidently want.

[Ritika Khurana]

David Radwin and Nick Cox Thank you both those suggestions worked for my dataset.