keeping one observation, when 2 or more variables repeat

[Cosmin Sandulescu]

Hi,

I am a beginner, trying to understand Stata and learning new things. I am working on a project, a retrospective cohort study design. I am trying to find out if an exposure to some medication does have protection on the exposed vs non-exposed population. When I got the data, they gave me each time when a diagnosis was made. I want to keep for the same id, just the first time the diagnosis was entered. I tried duplicates drop person dg, but not sure if that is the correct one and enough. There is also same person (persnumber) with multiple diagnosis /dg(0, 1, 2, ... 7).

Let me give you an example:

id persno dg date
1 2345 4 1/14/2011
2 2345 4 3/15/2017
3 2345 7 2/3/1988
4 2001 5 etc
5 2001 6
6 2017 0
7 2001 0

What I want, or how should it look like

id persno dg(0) dg(4) dg(7) dg(5) date
1 2345 0 1 1 0 1/14/2011 but also 2/3/1988 for dg(7)
2 2001 0 0 0 1
3 2017 1 0 1 0

I know that I could use tab dg, gen(dg_) which will give me dummy variables, but what to do when person duplicates, and also dg duplicates? and I would like to keep the first time the diagnosis was made, and how to connect second dg of interest to its date?

Thank you!

[Marcos Almeida]

Welcome to the Stata Forum / Statalist.

Please take some time to read the FAQ, particularly on how to share data.

Unfortunately, your query is not clear to me.

[Cosmin Sandulescu]

Thank you,

I see that my post is quite different, then how I wanted.

Initial Data

I know that I could use tab dg, gen (dg) to get dg(0) ....dg(7), but I do not know how to get rid off the duplicates dg() and persno, but keeping just the first duplicate whit the first date.

I also, need for the same persno with different dg, to create another row, but with the first date, too.

[Marcos Almeida]

It seems you wish to apply the - reshape wide - command. Unfortunately, I'm not with Stata at hand at this very moment.

That said, I fail to envisage the reason for reshaping wide, since the long format is preferred in most situations, more so under longitudinal data.

[Phil Bromiley]

I would look to sorts and lags to get what your original request asked. You can sort and then use a drop if condition to drop if there are repeated observations with the same diagnosis.