Help dropping observations and generating a new variable

[carla Oth]

Hi all,

I am having trouble with my MSc. Thesis. I have observations that look like this:

PersonID	Role	Year
A	1	2013
A	1	2014
A	1	2015
A	2	2013
A	2	2014
B	1	2014
B	1	2015

Basically, as it currently is, my data tells you what different jobs a person had and between what years. Therefore, A had 2 different jobs (1 and 2), 1 in years 2013-2015 and 2 in years 2013-2014. The problem is that I am not interested at all. I would like to have 1 observation per person, and I do not care about what different jobs/roles they had.

What I want to have is 1 observation per person, with the least recent year (for A this would be 2013), and one new variable that tells you the number of years that the person worked (i.e. the number of observations per PersonID). it would be like:

ID	Year	Experience_years
A	2013	5
B	2014	2

Do you think that is possible?

Thank you very much,

Carla

[Mike Lacy]

Documentation you would find useful would include -help system variables- and -help by-. Also, you should re-read the StataList FAQ to learn about using -dataex- to post example data.

// If there are any instances of multiple observations per person-year, the following will
// not give correct answers, so checking first is a good idea.
duplicates report PersonID Year
//
sort PersonID Year
by PersonID: gen Experience_years = _N
// Next command works because -sort- put the observation with the oldest year at the beginning.
by PersonId: keep if _n == 1

Note that the -keep- command retains only the oldest year's data on jobs etc. for each person, but I am taking you at your word that this is what you want.