Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: RE: Unbalancing a panel data set with country pairs

From   Matei Frunzetti <[email protected]>
To   [email protected]
Subject   RE: st: RE: Unbalancing a panel data set with country pairs
Date   Sat, 22 Jan 2011 13:57:27 +0100

Zitat von Nick Cox <[email protected]>:


. search identifier

points you to useful information. In this case,

egen idnew = group(id1 id2), label

is one of the easiest solutions. There is also a miniature review of possibilities at

SJ-7-4 dm0034 . . . Stata tip 52: Generating composite categorical variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox Q4/07 SJ 7(4):582--583 (no commands)
        tip on how to generate categorical variables using
        tostring and egen, group()

but the help to -egen- -- which I did refer you to earlier -- contains enough to solve this problem.

[email protected]

Matei Frunzetti

Thank you for your timely answer and excuse me for not being clear
enough: my main problem is that the country pairs have a single id for
each country, so i need to introduce a new id with its values being
unique identifiers for the respective country pairs such that i can
use by etc.

id1   id2    idnew  var1 ...

cty1  cty2   pair1  234
cty2  cty1   pair2  456

I am aware that this must seem pretty banal but i simply cannot find a
thread or help that would adress this problem, I probably just don't
know where to look.

Zitat von Nick Cox <[email protected]>:

I assume that "country pair" defines a panel identifier.

Look at the help for -egen-, which gives you one way to start.

In terms of missings,

egen missing_in_obs = rowmiss(varlist)
egen missing_in_panel = total(missing_in_obs), by(panelid)
drop if missing_in_panel

In terms of not enough years, you don't quite say what the time
resolution of your data is. But assuming you have yearly
observations then

bysort panelid : drop if _N < 17

drops the short panels. If it's half-years, quarters, months, etc.
adjust as needed.

You need to plug in your own <varlist> and <panelid>.

The big concept you appear to be missing is that of working -by:-.
There is a tutorial at

SJ-2-1  pr0004  . . . . . . . . . . Speaking Stata:  How to move
step by: step
        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
N. J. Cox
        Q1/02   SJ 2(1):86--102                                  (no
        explains the use of the by varlist : construct to tackle
        a variety of problems with group structure, ranging from
        simple calculations for each of several groups to more
        advanced manipulations that use the built-in _n and _N

[email protected]

Matei Frunzetti

I 'm working on a panel data set over 17 years. It's fairly unbalanced
und i need to drop all observations for country pairs that either lack
full length (as in years) or have missings in one of the independant
variables. The problem is that i have to delete all observations of
these country pairs for all years if only one or more variables have a
missing or if it is one or more years short. I ran into a dead end
trying to figure out how to imply the rest of the observations of a
"faulty" country pair into the drop command.

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index