Generate one exit variable based on different variables

[Morten Hans Jensen]

Hi Statalist

I want to generate an exit (out of cohort) variable. I have a large dataset with a lot of different dates.
You can exit the cohort when you turn 5 (60months), move (move) or die (death). Variable name in ().

I am only interested in seeing the dates for these outcomes in the interval of 1jan2010 to 1jan2011. The others should just be missing values.

My try that does not work as I still have dates outside of the interval:

gen exit = .

{
replace exit = min(death, move, 60months) if inrange(death, td(1jan2010), td(1jan2011)) | inrange(move, td(1jan2010), td(1jan2011)) | inrange(60months, td(1jan2010), td(1jan2011))
}

I do not have access to the data at the moment, sorry. I am looking for suggestions on optimizing my code or a better way of doing it.

[Morten Hans Jensen]

Anybody with a suggestion?

[Nick Cox]

I think I need to see a data example. https://www.statalist.org/forums/help#stata explains -- including an explanation that fake data with the same structure are fine if your data are confidential.

For one, 60months can't be a variable in your dataset. The name is illegal.

[Morten Hans Jensen]

Id	Move	Death	Fiveyears
1	.	.	07-08-2010
2	.	.	20-11-2010
3	15-02-2015	.	21-10-2010
4	.	.	07-05-2011
5	.	06-05-2010	14-04-2011
6	.	.	05-03-2011
7	16-04-2016	.	16-04-2012
8	07-06-2010	.	21-05-2011
9	.	18-11-2010	22-02-2012

Hi Nick, thank you for taking a look.
I have made a data example.

I am interested in getting a variable that has the date for those that move, die or turn 5 within 1jan2010 to 1jan2011. Death can only be in this interval as it is a variable I have generated. The other variables can be outside this.
I can solve it but the solution I have come up with is labor "heavy" and seems overcomplicated.

[Nick Cox]

OK. that's a good start but what are these variables? Strings with dates? Numeric variables with a daily date display format? What are the blanks supposed to be? Empty strings or numeric missings? Using dataex -- see link in #3 -- would clarify these points.

[Morten Hans Jensen]

The variables are dates for when an event occurs.
Blanks are empty string i.e. an event has not occured.

The data in the table is how I would see it. I do not have access to the data at the moment.

[Nick Cox]

When I copy and paste, I see not blanks but system missings in the (nearly) empty cells. So I surmise this is what you have and what you want.

The use of a local macro is just to reduce my typing.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte Id float(Move Death Fiveyears)
1     .     . 18481
2     .     . 18586
3 20134     . 18556
4     .     . 18754
5     . 18388 18731
6     .     . 18691
7 20560     . 19099
8 18420     . 18768
9     . 18584 19045
end
format %tdDD-NN-CCYY Move
format %tdDD-NN-CCYY Death
format %tdDD-NN-CCYY Fiveyears

local range mdy(1,1,2010), mdy(1,1,2011)
gen exit = inrange(Move, `range') | inrange(Death, `range') | inrange(Fiveyears, `range')

list 

     +--------------------------------------------------+
     | Id         Move        Death    Fiveyears   exit |
     |--------------------------------------------------|
  1. |  1            .            .   07-08-2010      1 |
  2. |  2            .            .   20-11-2010      1 |
  3. |  3   15-02-2015            .   21-10-2010      1 |
  4. |  4            .            .   07-05-2011      0 |
  5. |  5            .   06-05-2010   14-04-2011      1 |
     |--------------------------------------------------|
  6. |  6            .            .   05-03-2011      0 |
  7. |  7   16-04-2016            .   16-04-2012      0 |
  8. |  8   07-06-2010            .   21-05-2011      1 |
  9. |  9            .   18-11-2010   22-02-2012      1 |
     +--------------------------------------------------+

[Morten Hans Jensen]

Thanks for the reply Nick.