How to remove the entire data for a certain firm if it starts at certain date

[Yousef Kaddoura]

hello!

I want to ask a small question, I want to remove all firms that have the date 1962. The reason Why I want to do this is because I'm trying to do a robustness test where if I remove a subset of my sample, I should get the same results.

A small sample from my sheet follows,

datayearfiscal firm
1970 1000
1971 1000
1972 1000
1973 1000
1974 1000
1975 1000
1976 1000
1977 1000
1962 1001
1963 1001
1964 1001
1965 1003
1967 1004
1968 1004
1969 1004
1970 1004

What It should end up like:

datayearfiscal firm
1970 1000
1971 1000
1972 1000
1973 1000
1974 1000
1975 1000
1976 1000
1977 1000
1967 1004
1968 1004
1969 1004
1970 1004

You could see the entire firm was deleted based on that they had the year "1962" this is just an example! not the actual data collected!

Regards And thank you!

[Yousef Kaddoura]

Hello,

I figured it out!

What I Did is first running the following code:

 
 replace dateyearfiscal = . if dateyearfiscal < 1963

this will replace all 1962 points with ".".

Afterwards I ran the following

 
 bysort firm (dateyearfiscal) : drop if missing(dateyearfiscal[_N])

this simply deletes the entire panel.

Now this won't work every time because you might have a lot of missing data and you just want to delete the panel if only the first point is missing. so i thought of running such a code

 
 bysort firm (dateyearfiscal) : drop if missing(dateyearfiscal[1])

Feel free to correct me, but this was my conclusion and option 1 worked for me since i had no gaps ^_^

Regards,

[Jean-Claude Arbaut]

If you want to remove all firms having any record for 1962 (not necessariliy the first), you can do that:

gen u=datayearfiscal==1962
bysort firm: egen v=max(u)
drop if v
drop u v

Or alternatively:

gen u=datayearfiscal==1962
bysort firm (u): drop if u[_N]
drop u

If, however, you want to remove only firms for which the first record (in date order) is for 1962, then one line is enough:

bysort firm (datayearfiscal): drop if datayearfiscal[1]==1962

Replacing dateyearfiscal with a missing is a possibility, but it will fail if there is missing data in this variable, and you don't want to remove those firms. You could instead replace with a small value (say, 0, which won't be a valid year nor a missing value). However, using an additional indicator variable looks cleaner to me.

[Yousef Kaddoura]

What said Jean works too ^_^

Thanks!

[Yousef Kaddoura]

Yes Jean, Your way actually works in more cases and mine was just a special case I guess. I did benefit from your codes because I had to do this for variables that had more than 1 missing value as well. Much Appreciated.