Unstacking Pairs

[Luis Pecht]

Hi I need a small data management trick that i am assuming it is easier in Mata than Stata.

assume we have that stacking format:

clear
input int(i j) byte similarity
 9  43 25
28  67 21
28  43 26
89 124 21
89 160 21
89 123 21
89 167 21
89 125 22
89 161 22
89 154 22
89  90 22
89 157 22
89 162 23
end

expected output Mata Matrix would be:

first column = unique i
follows columns be the J pair, but descendant ordered by similarity from left to right

9	43
28	43	67
89	162	157	90	...

thks in advance.

[Mike Lacy]

The result you show can be done with Stata commands, and the resulting data structure can be made into a Mata matrix:

clear
input int(i j) byte similarity
 9  43 25
28  67 21
28  43 26
89 124 21
89 160 21
89 123 21
89 167 21
89 125 22
89 161 22
89 154 22
89  90 22
89 157 22
89 162 23
end
//
// Order j within i by similarity, and capture this order in a variable.
bysort i (similarity): gen int simord = _n
drop similarity
// Get the wide layout you want.
reshape wide j, i(i) j(simord)
list
// Assuming you want a Mata matrix as an end in itself.
mata X = st_data(., .)
mata X