Possible to select only unique observations (without dropping duplicates)?

[Stine Nielsen]

I'm working on a dataset with about 45,000 observations for about 5500 unique id's.
Is there a way to do e.g. tab of only the observations with a unique id? I know I can drop duplicates, but I need them later. However for some of my analysis I only want to display the observations that have a unique id.
Any thoughts?
Thank you!
ps I work in Stata 13.1/IC on Mac.

[Attaullah Shah]

You have not told us about the purpose or about your data set. An example of data set using dataex from SSC will help us in pinpointing what you want to accomplish. Anyway, there are several alternatives of what you want to dol;

• First one is to preserve your data set,
• then drop duplicates
• then restore.

lternatively, you can drop duplicates, save a temporary file, and then reload the original file for other uses.

[Nick Cox]

On using the term distinct rather than unique, see early comments in http://www.stata-journal.com/sjpdf.h...iclenum=dm0042 The whole paper may be useful too.

I think what you want is along these lines

egen tag = tag(id) 
... if tag 

which works with just one example (the first occurrence) of each distinct identifier, so that for example

list id if tag, noobs

shows the distinct identifiers just once each.

[Stine Nielsen]

Thank you. I am an epidemiologist working on infectious diseases. My id var refers to individuals and my observations are e.g. about when people got tested (and many got testet multiple times).
I will try with preserve and restore. Was hoping it would be possible with an "if command" something like if id == "unique" - but I guess that is not possible.

Kind regards from Madrid, Spain.

Stine Nielsen

Phd student, EPIET alumni and freelance epidemiologist
LinkedIn: www.linkedin.com/in/stinenielsenepi
Twitter: www.twitter.com/StineNielsenEPI

[Nick Cox]

I think you're replying to Attaullah in #2. It really is possible with an if qualifier (not command).

[Stine Nielsen]

Thx Nick! Yes I replied to post#2 before reading your reply.

egen tag = tag(id) 
tab sex if tag==1

works for me.
Thanks again.

[Stine Nielsen]

Is there a way to make sure that I always tag the most recent observation per id?

I tried this, but it didn't work:

bysort id (datevar) : egen tag2 = tag(id)

Each id has many observations and I want to choose / tag the observation with the most recent date

max(datevar)

per id.
Any tips on how to do that?

[Nick Cox]

 
 bysort id (datevar) : gen byte is_last = _n == _N

[Stine Nielsen]

Almost works. But I need it to ignore missing values in datevar - I tried adding

if datevar!=.

at the end of your suggested code - but this didn't work well (gave me several id's where the observation with the last datevar was not included in "is_last".
More tips?

[Nick Cox]

gen ismissing = -missing(datevar) 
bysort id (ismissing datevar) : gen byte is_last = _n == _N

Note particularly the minus sign.

missing(datevar) is 1 if the argument is missing and 0 otherwise. If we negate that (-1 and 0), then missings are sorted first for each individual.

[Stine Nielsen]

Fantastic! Thanks so much!