Panel data set seperating IDs based on a variable for subsample analysis

[krishantha Ainsworth]

Der Stata list,


I need to separate IDs of a panel dataset for subsample analysis based on values they have for a perticular variable ( say for an example you want to seperate goods based on price increases and decreases).
An example of what I need is below. (

Year	ID	Variable
1990	1	1
1990	2	23
1990	3	5
1990	4	-4
1990	5	0
1990	6	.
1991	1	4
1991	2	6
1991	3	0
1991	4	45
1991	5	0
1991	6	0
1992	1	0
1992	2	3
1992	3	-3
1992	4	0
1992	5	0
1992	6	.
Criteria
1) see what IDs have (at least ) one positive value throughout the panel (e.g IDs 1,2,3,4)
2) see what IDs have (at least ) one negative value throughout the sample (e.g IDs 4 and 3)
3) separate (create sub samples) above 1) and 2) IDs by providing them separate identifiers (probably by creating another variable) I appreciate if you could help me

[Øyvind Snilsberg]

egen crit1 = max(variable>0 & variable<.), by(id)
egen crit2 = max(variable<0) ,by(id)
egen group = group(crit1 crit2)

[krishantha Ainsworth]

Thank you so much Øyvind . One other question I have is how can I run a regression for both groups, those who increase and decrease ? If I run it as

 regress y x if variable < 0 (or variable >0)

does it satisfy my objective ? On the other hand what if
I run them for if crit1/crit2 ==1?

[Øyvind Snilsberg]

to restrict the sample to those who increase and decrease,

regress y x if crit1==1 & crit2==1

or, equivalently,

regress y x if group==3