long type data_salary differences

[TaeWon]

[Very sorry, text version can't figure it out the exact column. Please look at the photo file. I'm so sorry.]

I have a long type data like following;

hid pid rel sex marriage salary
1 101 1 1 1 300
1 102 2 0 1 0
2 201 1 1 1 0
2 202 2 0 1 250
2 203 3 1 0 0
3 301 1 1 1 350
3 302 2 0 1 450

where if rel=1 then husband, rel=2 wife, rel=3 children, and if sex=1 then male, sex=0 female.
In this case how can I get the differences between husband's and wive's salary, like following ;

differences
hid pid rel sex marriage salary A_type B_type C_type
1 101 1 1 1 300 300 300
1 102 2 0 1 0 -300 -300
2 201 1 1 1 0 -250 -250
2 202 2 0 1 250 250 250
2 203 3 1 0 0 0
3 301 1 1 1 350 -100 -100
3 302 2 0 1 450 100 100

The kind of types do not matter.

[TaeWon]

V\ery sorry, text version can't figure it out the exact column. Please look at the photo file. I'm so sorry.

[Carlo Lazzaro]

TaeWon,
unfortunately, no photo file is attached to your post.
Anyway, I would take a look at -help reshape-.
PS: just out of curiosity, what is the meaning of *_type (*=different uppercase letters) in yuor target database?

[TaeWon]

Thanks for your attention.
Actually photo file is posted on your 'upper reply'.
I'm sorry but I'm novice at the forum.
If you see the photo file you can figure it out the *_type immediately.

Kind regards, TaeWon

[Nick Cox]

 
clear 
input hid pid rel sex marriage salary
1 101 1 1 1 300
1 102 2 0 1 0
2 201 1 1 1 0
2 202 2 0 1 250
2 203 3 1 0 0
3 301 1 1 1 350
3 302 2 0 1 450
end 
gen married = inlist(rel, 1, 2) 
bysort hid married : gen diff = salary[3 - _n] - salary if married & _N == 2

Technique is documented at http://www.stata-journal.com/sjpdf.h...iclenum=dm0043

[TaeWon]

Thanks Nick!
It's awesome!