The code looks good. Perhaps the data is not quite as it should be.
If you want to keep all the data, then you can still segregate the
spouses.
In your family of 5, I guess this
spouse1 spouse2
1 1 .
2 . .
3 . 3
4 . .
5 . .
gen not_spouse = !((spouse1 < .) | (spouse2 < .))
bysort site family (not_spouse) : gen partner_income = personal[3 - _n]
if !not_spouse
There are two states to not_spouse. not_spouse is true (1) if somebody
is not a spouse and false (0) if somebody is.
Sorting on -non_spouse- within a family sorts all the 0s to the first
observations in each family.
Your set-up may differ, but something like this should still be
workable.
Nick
[email protected]
Ekaterina Selezneva
for me your solution with "[3 - _n]" is the most elegant and most
easily understood.
But how to handle the situation if I have data on all the household
members, and let's say, I want to keep them all. I have two variables
identifying a couple within each household: "spouse1" and "spouse2"..
so, if
there is a household composed of 5 members, and N1 and N3
are married, then spouse1=1 and spouse2=3.
I would do something like
#delimit;
bysort site censusd family : gen partner_income = personal[spouse1]
if person==spouse2;
bysort site censusd family : replace partner_income =
personal[spouse2] if person==spouse1;
it works fine if both partners are present in the sample, but if only
one is in, the partner_income variable is replaced sometimes with the
individual
income of the individual and sometimes with the missing value.
What could be wrong with the code?
Thursday, May 15, 2008, 4:24:34 PM, you wrote:
> Using the same assumption of a Noah's ark situation in which everyone
is
> paired off, with no dependents, in-laws, etc., there are other minor
> variations on Scott's theme.
> bysort site family: gen partner_income = cond(_n == 1, personal[2],
> personal[1])
> bysort site family: gen partner_income = personal[3 - _n]
> I am fond of the second. If it looks puzzling, just go through the two
> cases. If _n is 1, then 3 - _n is 2, and vice versa.
> The key underlying principle, if it is not familiar, is that under
-by:-
> _n is interpreted within groups, not within the entire dataset. There
is
> a leisurely tutorial
> at
> <http://www.stata-journal.com/sjpdf.html?articlenum=pr0004>
> N.B. this is in the public domain.
> Nick
> [email protected]
> Scott Merryman
> What defines a partner? Someone within the same two-person family?
> Perhaps something like this:
> clear
> input site family person personal_labor_income
> 1 1 1 2000
> 1 1 2 2300
> 1 2 1 200
> 1 2 3 3000
> 2 10 4 3400
> 2 10 5 3500
> end
> bysort site family : gen partner_income = cond(_n ==1,
> personal[_n+1],personal[_n-1])
> Ekaterina Selezneva
>> I have a dataset with some information on a sample of married
>> couples. For identifying a single person, one needs to know a
>> "site"-number, "family"-number withing the site, and then the
>> "person"-number within the family. As this is a subsample of some
>> bigger dataset, so not all sites/families/persons are presented in
>> it. Let's say, something like:
>>
>> site family person personal_labor_income
>> 1 1 1 2000
>> 1 1 2 2300
>> 1 2 1 200
>> 1 2 3 3000
>> 2 10 4 3400
>> 2 10 5 3500
>>
>> THE PROBLEM: I need to create a variable containing the "personal
>> labor income" of partner.
>>
>> Unfortunately, I've spent a day, and havn't
>> succeded to solve this seemingly simple problem. I will be grateful
> for
>> any hints.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
