Create rank variable for regression analysis

[Rochelle Zhang]

Dear Statalisters,

I want to run a panel data regression S on x1,x2, x3, x4. Following prior research, I want to convert all continuous variables to 0-99 percentile ranks.

Panel id: firmid. Time id: year

What I did:

bysort year: egen srank=rank(S)
bysort year: egen x1rank=rank(x1)

sum srank ranges from 1 to 1144

my sample has around 25000 observations.

Question: how do I convert the ranks of all my variables to 0-99 percentile?

Thanks,
Rochelle

[Nick Cox]

FAQ

http://www.stata.com/support/faqs/st...ons/index.html

[Rochelle Zhang]

yes, Nick. I actually found your FAQ before I post to the forum. that is why I used egen. I got ranks, but they are not bounded in 0-99 range.

I hope this is not a stupid question: is it possible to make all the ranks within 0-99 range, that is after I
bysort year: egen srank=rank(S)
bysort year: egen x1rank=rank(x1)

can I make srank and x1rank both bounded in 0-99 range.

[Nick Cox]

This really is answered in the FAQ cited. I draw your attention to the section on plotting positions ending

All of these examples return results between 0 and 1; for percentages, just multiply by 100.

[Rochelle Zhang]

You are absolutely right. Thanks you Nick !

Best wishes,
Rochelle