Count number of variable combinations

[Maarten Huurne]

I want to create a variable that counts the number of combinations between two variables. I have a variable 'buyerID' and 'sellerID' for 100,000 transactions. I want to know how many times a certain buyer and seller have done business together.

Thank you in advance!

[Rich Goldstein]

here is one way:

egen newvar=group(buyerID sellerID)

you may want the label (sub)option - read the help file
you can then just

ta newvar

[Carlo Lazzaro]

Maarten:
you may want to try something along the following lines:

set obs 3
g Buyer=1
g Seller=_n+1 in 1/2
replace Seller=3 in 3
egen intersection=group(Buyer Seller)
bysort Buyer Seller: count if intersection>=1

[Carlo Lazzaro]

Maarten:
you may want to try something along the following lines:

set obs 3
g Buyer=1
g Seller=_n+1 in 1/2
replace Seller=3 in 3
egen intersection=group(Buyer Seller)
bysort Buyer Seller: count if intersection>=1

[Nick Cox]

Sounds like

 
egen group = group(buyer seller) , label

except that "do business" might mean also same partners, reversed roles. If so, see http://www.stata-journal.com/sjpdf.h...iclenum=dm0043

[Maarten Huurne]

Thanks everybody for the valuable advice!

[Maarten Huurne]

I have an additional question. With the answers provided, the codes count the number of combinations. Additionally, I want to count the number of each combination. So, the combination '1 2' '1 2' should be counted as 2. I have enclosed an image of my dataset for more clarity.

[Nick Cox]

bysort group: gen freq = _N

Consider also (destructive reductions)

contract group
rename _freq freq 
contract freq

[Maarten Huurne]

Thanks for the solution!