Help with ties when using egen and the rank option

[m_burke]

Let's say I have two variables "score" and "rank." I would like "rank" to be a rank of each score allowing for ties. That is, if a tie occurs, each "score" should receive the same rank and the forthcoming rank should "skip" ahead by the number of ties that preceded it (see below for an example).

I have tried using the egen function with the rank option, but I could not get it to produce what I would like.

Score	Rank
20	1
21	2
22	3
22	3
23	5
24	6
24	6

[Nick Cox]

You need to use the egen command (not function) with the rank function (not option) and the track option. It's documented, as is the reverse field option.

Perhaps the names aren't good. In my youth, people could tie at 2nd equal with the same time in a track event or the same height or distance in a field event. That seems rarer now with automatic timing, photofinishes and so forth. I spent some time looking for names for these in 1999 when Richard Goldstein and I first implemented them unofficially, but couldn't find any names in the literature and ended up inventing my own. At least that's my story: Richard may have a different one.

Since then I have only come across schoolmaster's rank as an alternative name for the field option, which I won't try to persuade anyone to use. There was a comment in http://www.stata-journal.com/sjpdf.html?articlenum=pr0046

[Clyde Schechter]

egen Rank = rank(Score), track

does exactly what you've described.

[Dirk Enzmann]

Did you read the entry "rank" using help egen? If you don't understand by reading try out all options:

clear
input score
20
21
22
22
23
24
24
end
egen rank = rank(score)
egen rank_f = rank(score), field
egen rank_t = rank(score), track
egen rank_u = rank(score), unique
list, sepby(score)

By the way: The Statalist FAQ asks you (#3) to read the Stata online help before posting and (#6) to use your full real name (if you think that's pedantic - it helps to keep up the quality of this forum). Use "contact us" at the bottom right to ask the administrators to change your name.

[Nick Cox]

Note that even if a egen function didn't exist, this is something you could get directly:

sysuse auto 
bysort foreign (mpg) : gen rank = _n
bysort foreign mpg (rank) : replace rank = rank[1]

[Nick Cox]

I am concerned that you messed up your data with your use of round(). That function can't perform accurate decimal rounding of fractions because no function can do that. This follows from the fact that at bottom all arithmetic in Stata is binary.

Also, you seem to be confusing display formats with how data are stored.

On the contrary, the egen function rank() treats equal values equally.