Numbering new values of a variable 1 to N

[Josephine George]

My data has a country variable using three digit ISO-3166 codes. The underlying codes aren't quite sequential anyway, and then because I'm using a subset of countries there are gaps. To make some loops run smoothly, I would like to have a variable where countries are 1, 2, 3 etc. I think the backbone of the code I need is something like:

//sort country wave id
// gen newvar=1 if obs==1
//gen newvar=newvar(_n-1) if country(_n-1)=country
//replace newvar=newvar(_n-1)+1 if country(_n-1)!=country

I expect there are obvious pitfalls to that approach that I'm missing, and this feels like something egen should be able to do in a single line. Any guidance appreciated.

[Joro Kolev]

Try

egen cntry = group(country)

[Joro Kolev]

Or alternatively if country identifies an observation

gen cntry = _n

and if none of those work provide a data sample with -dataex-.

[Nick Cox]

The idea is good, and can be coded up (You need == consistently when testing for equality.)

bysort country : gen newvar = 1 if _n == 1 
replace newvar = sum(newvar)

or

sort country 
gen newvar = sum(country != country[_n-1] )

but it's what

egen newvar = group(country) , label

does any way. (I added label as good practice going beyond you did.)

See FAQs

https://www.stata.com/support/faqs/d...p-identifiers/

https://www.stata.com/support/faqs/d...-with-foreach/

The latter FAQ is just a start on an enormous topic. Getting groupwise results doesn't always require loops; see e.g. statsby, runby (SSC), rangestat (SSC).

It's a fair bet that what you want to do next is covered by hundreds of threads here, so do make it specific.

[Josephine George]

egen group worked perfectly. Thanks!
Appreciate the pointer to the FAQ too -- no doubt it will save me posting more obvious questions here.