Retaining the entire ID if there are certain variables in the panel data

[Kim JongBin]

Good Morning world

English is not my language so google help me. Please understand me.

I have stata 15.

I`m using panel data with 191 variables and this data has been investigated for about 22years

ID goes from 101 to 102, 103...... like this.

wave is time variable

Please check the attached file for the data shape.

Here`s my question.

If var1 has 2 or 3, I'd like to keep all the IDs.

If var1 doesn't have 2 or 3, I want to get rid of all of that ID.

So I wrote down like

gen nid=ID if var1 == 2 | var1 == 3
gen nnid =1 if nid < .

I don't know what kind of logic to proceed from now on.

Using loop(ex.forvalues) and "keep if" don't seem to work properly.

gen nid=ID if var1 == 2 | var1 == 3
gen nnid =1 if nid < .
keep if nnid == 1

OR

gen nid=ID if var1 == 2 | var1 == 3
gen nnid =1 if nid < .
forvalues wave = 1/22{
    keep if nnid == 1
    }

It`s not work and I know why this code doesn`t bring me to answer

But I don`t know how do i solve this case.

Please give me an advice on what logic to proceed with.

I hope this article followed the Advice on posting to Statalist.

Thankyou

[Hong Il Yoo]

Could you confirm if the following delivers what you're after?

gen double yes23 = [var1 == 2 | var1 == 3]
sort ID, stable
by ID: egen double min_yes23 = min(yes23)
drop if min_yes 23 == 0
drop yes23 min_yes23

[Nick Cox]

Hong Il Yoo gave a helpful answer, but the code can be slimmed down:

bysort ID: egen double min_yes23 = min(inlist(var1, 2, 3))
drop if min_yes 23 == 0
drop min_yes23

The general problem I call the creation of "any" variables -- any observations in a group match a criterion? -- and "all" variables can be discussed alongside.

FAQ https://www.stata.com/support/faqs/d...ble-recording/

Before Googling wildly, scanning the Stata FAQs is a good idea. Usually, somebody else had the same problem earlier!

[Kim JongBin]

Nick Cox Thank you for your advice specially FAQ url. I'll make good use of it.

[Kim JongBin]

Could you confirm if the following delivers what you're after?

gen double yes23 = [var1 == 2 | var1 == 3]
sort ID, stable
by ID: egen double min_yes23 = min(yes23)
drop if min_yes 23 == 0
drop yes23 min_yes23

Hong Il Yoo Thank you for your answer.

But It didn't work as I wished.

When I used your code some ID was deleted even it has 2or3

and some ID`s wave was deleted when I used your answer

the remaining ID is very different from when I used "drop if"

That is my big problem

I'll keep looking for solutions. Please let me know if you have a good idea.

It will be of great help to me. Have a good day

[Ali Atia]

I believe what you want is:

bysort ID: egen double max_yes23 = max(inlist(var1, 2, 3))
drop if max_yes23 == 0
drop max_yes23

Which is adapted from @Nick Cox's code in #3, but with max(inlist(var1,2,3)) instead of min(inlist(var1,2,3)), in order to set max_yes23 to 1 if there is at least 1 occurrence of 2 or 3.

[Hong Il Yoo]

Hong Il Yoo Thank you for your answer.

But It didn't work as I wished.

When I used your code some ID was deleted even it has 2or3

and some ID`s wave was deleted when I used your answer

the remaining ID is very different from when I used "drop if"

That is my big problem

I'll keep looking for solutions. Please let me know if you have a good idea.

It will be of great help to me. Have a good day

As far as I can tell, the only issue with my code example is that I typed -drop if min_yes 23 == 0- when I should have said -drop if min_yes23 == 0-. At least for now, I don't see how it is possible that the code drops some waves and replaces the content of your ID variable. If you can share with us a subset of your data that can be used to replicate the problems that you have mentioned, one of us may be able share a more useful suggestion.

[Hong Il Yoo]

Ah yes Ali Atia's observation is spot on, it should have been -max- rather than -min-.

[Nick Cox]

I am corrected too: in essence "any" is returned by taking the maximum over an (0, 1) indicator and "all" by taking the minimum.

[Kim JongBin]

I believe what you want is:

bysort ID: egen double max_yes23 = max(inlist(var1, 2, 3))
drop if max_yes23 == 0
drop max_yes23

Which is adapted from @Nick Cox's code in #3, but with max(inlist(var1,2,3)) instead of min(inlist(var1,2,3)), in order to set max_yes23 to 1 if there is at least 1 occurrence of 2 or 3.

Ali Atia

Your code is working now!

When I used Ali's code, all waves of an ID with any observation 2 or 3 were marked as 1 in max_yes23.

The previous code, min_yes23, had only a partial wave of IDs with 2 or 3.

And Nick Cox and Hong Il Yoo The advice you gave me was a great help in my study of Stata

Your advice will be useful next time.

All the advice you gave me became a big study. Thanks. I really appreciate it.

I wish you all the best.