Drop panel-observations if data have no count

[Jana Schue]

Hi everyone

I have the variables "year" (2009-2019) for each "company key" (e.g.: 1661) and two count variables (the other two columns) and want to drop all observations of a "company key" if every "year" per "company key" has no number.
If a "company key" has a number in one "year", I want to keep every observations of the "company key".

As you can see in the picture:
1661 has no count and needs to be dropped.
1690 has a count in 2010 and every observation should be kept.

Thank you for your help!

Best,
Jana

[Fei Wang]

Jana, if the last two columns are missing or non-missing at the same time, then you may run the following codes.

bysort companykey: egen tempvar = total(lastcolumn), missing
drop if tempvar == .
drop tempvar

[Nick Cox]

bysort gvkey (whatever) : drop if missing(whatever[1])

If after sorting the first value for each company is missing, then they all are.

If you need that both variables should always be missing, you need a two-step:

bysort gvkey (something) : gen todrop = missing(something[1])
bysort gvkey (whatever) : drop if todrop & missing(whatever[1])

Examples using dataex are greatly superior (please) to images, deprecated at FAQ Advice #12

[Nick Cox]

#2 needs a fix. The total of several missings is 0, not missing, unless you specify missing as an option.

[Jana Schue]

Nick Cox!
Thank you very much.

And also thanks for the hint, I will do that next time!

Best,
Jana

[Fei Wang]

#2 needs a fix. The total of several missings is 0, not missing, unless you specify missing as an option.

Thanks, Nick, for pointing this out. I have corrected #2 codes. And your solution is much neater.

[Nick Cox]

Fei Wang Thanks!

https://en.wikipedia.org/wiki/On_the...you%27re_a_dog

On the internet everyone knows you're a cat.... .