Creating a dummy variable for each observation separately

[Sebastian Finke]

Hello,

I have a nummeric variable which is represented in % and I need to recode this variable. To be specific, I need to have for each observation that has a value >0 a different natural number. The goal is to create a fixed effect dummy for a regression analysis. So, for instance:

Observation 1: value 5 --> obtains value 1
Observation 23: value 3 --> obtains value 2
Observation 58: value 9 --> obtains value 3
...etc

Is there a command that can do that automatically?

Thanks in advance!

[Carlo Lazzaro]

Sebastian:
do you mean something along the following lines?

. set obs 3
number of observations (_N) was 0, now 3

. g A=runiform()/100

. egen flag=group(A) if A>0

. list

     +-----------------+
     |        A   flag |
     |-----------------|
  1. | .0034887      3 |
  2. | .0026689      2 |
  3. | .0013665      1 |
     +-----------------+

[Sebastian Finke]

Hello,

thanks for your answer. I tricked around after a bit of waiting and I found a solution: I made an "inverse"-string command and then used the encode command to get the values I wanted. Sorry for bothering!

Have a nice day!

[Carlo Lazzaro]

Sebastian:
thanks for sharing your solution.