Indicate a change in a variable between two periods

[David Dahlgren]

Hi,
I have a problem were I want to identify if a variable from an individual change between 2 periods (t - t+1). The variable that I want to observe is if a houseID change. The maximum number of years are 10 but it´s not a balanced sample, were some individuals only have a value 1 period. The data is panel data.

The problem I have with my code is that independently if there is a change or not, the data indicates a 1 in the last period for all individuals. I don´t know how to solve this. Bellow is the code that I use.





gen change = 0
bysort individualID (year): replace change = 1 if houseID!=houseID(_n+1)


Best
David

[Maarten Buis]

For the last observation you obviously cannot know whether the non-existent next observation has the same houseID or not. So the correct value should be missing.

bysort individualID (year) : gen change = 0 if _n != _N
bysort individualID (year): replace change = 1 if houseID!=houseID(_n+1) & _n != _N

[David Dahlgren]

Thanks for the reply Maarten, and you are off course right with the missing value. I guess that is what the last part does, if _n!= _N. I am a bit unfamiliar with the use of n and N. Say that we would like the first period to have a missing values, how would we code it then? For example a set up were the specification is houseID!=houseID[_n-1]

Also, just so I did it correctly, when I used ( ) I got a error message "unknown function houseID( )", while changing to [ ] it worked fine. Is that right?


Best
David