. * Tutorial 20: Programming with Mata: from Stata networks to Mata and back . * Requires the core nwcommands. . . * 1. A network as a Mata matrix . nwwebuse florentine . nwtomata flomarriage, mat(A) . return list macros: r(adj_copy) : "A" r(adj) : "(*__nw.nws.pdefs[2]->get_matrix())" r(netobj) : "__nw.nws.pdefs[2]" r(netname) : "flomarriage" . mata: rows(A), cols(A) 1 2 +-----------+ 1 | 16 16 | +-----------+ . mata: rowsum(A)' 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 +---------------------------------------------------------------------------- 1 | 1 3 2 3 3 1 4 1 6 1 3 0 3 2 4 +---------------------------------------------------------------------------- 16 ------+ 1 3 | ------+ . mata: A[1..4, 1..4] [symmetric] 1 2 3 4 +-----------------+ 1 | . | 2 | 0 . | 3 | 0 0 . | 4 | 0 0 0 . | +-----------------+ . . * A copy: changing it does not change the network . mata: A[2,1] = 7 . nwtomata flomarriage, mat(B) . mata: A[1..3, 1..3], B[1..3, 1..3] 1 2 3 4 5 6 +-------------------------+ 1 | . 0 0 . 0 0 | 2 | 7 . 0 0 . 0 | 3 | 0 0 . 0 0 . | +-------------------------+ . mata: mata drop A B . . * 2. A measure of our own, as a small Mata function . mata: ------------------------------------------------- mata (type end to exit) -------- : real colvector nw_reach2(real matrix A) > { > real matrix B > B = A > _diag(B, 0) // no self-ties; the diagonal arrives as missing > B = ((B + B*B) :> 0) // tie, or a common partner > _diag(B, 0) > return(rowsum(B)) // how many others are within two steps > } : real colvector nw_nbrmean(real matrix A, real colvector x) > { > real matrix B > B = A > _diag(B, 0) > return((B*x) :/ rowsum(B)) // mean of x over the neighbours (. for an isola > te) > } : end ---------------------------------------------------------------------------------- . . nwtomata flomarriage, mat(A) . mata: nw_reach2(A)' 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 +---------------------------------------------------------------------------- 1 | 6 10 9 8 6 3 9 4 11 2 6 0 11 7 8 +---------------------------------------------------------------------------- 16 ------+ 1 10 | ------+ . mata: x = st_data(., "wealth") . mata: nw_nbrmean(A, x)' 1 2 3 4 5 +----------------------------------------------------------------------- 1 | 103 47.66666667 61.5 67.66666667 83.33333333 +----------------------------------------------------------------------- 6 7 8 9 10 ----------------------------------------------------------------------- 1 36 42.5 8 31 10 ----------------------------------------------------------------------- 11 12 13 14 15 ----------------------------------------------------------------------- 1 70 . 99 75.5 35 ----------------------------------------------------------------------- 16 ---------------+ 1 46 | ---------------+ . . * 3. Back to Stata: a node variable, a matrix, a new network . mata: st_store(., st_addvar("double", "reach2"), nw_reach2(A)) . mata: st_store(., st_addvar("double", "nbrwealth"), nw_nbrmean(A, x)) . list _nwnode wealth reach2 nbrwealth in 1/6 +------------------------------------------+ | _nwnode wealth reach2 nbrwealth | |------------------------------------------| 1. | acciaiuoli 10 6 103 | 2. | albizzi 36 10 47.666667 | 3. | barbadori 55 9 61.5 | 4. | bischeri 44 8 67.666667 | 5. | castellani 20 6 83.333333 | |------------------------------------------| 6. | ginori 32 3 36 | +------------------------------------------+ . mata: st_matrix("R", nw_reach2(A)) . matrix list R R[16,1] c1 r1 6 r2 10 r3 9 r4 8 r5 6 r6 3 r7 9 r8 4 r9 11 r10 2 r11 6 r12 0 r13 11 r14 7 r15 8 r16 10 . . mata: _diag(A, 0) . mata: P = A*A . mata: _diag(P, 0) . _nwsyntax flomarriage . nwset, mat(P) name(paths2) labs(`labs') undirected . nwsummarize paths2 -------------------------------------------------- Network name: paths2 Network id: 3 Directed: false Valued: true Two-mode: false Nodes: 16 Selfloop: false Edges: 43 Minimum value: 0 Maximum value: 2 Density: .358 Mean degree: 5.38 Temporal: false . nwtomata flomarriage, mat(A) . mata: _diag(A, 0) . mata: R2 = ((A + A*A) :> 0) . mata: _diag(R2, 0) . nwset, mat(R2) name(floreach2) undirected . nwsummarize floreach2 -------------------------------------------------- Network name: floreach2 Network id: 4 Directed: false Valued: false Two-mode: false Nodes: 16 Selfloop: false Edges: 55 Minimum value: 0 Maximum value: 1 Density: .458 Mean degree: 6.88 Temporal: false . mata: mata drop A P R2 x . . * 4. A reusable command . capture program drop nwreach2 . program nwreach2, rclass 1. syntax [anything(name=netname)], GENerate(name) [NETwork(name)] 2. _nwsyntax `netname' 3. _nwdatasync `netname' 4. confirm new variable `generate' 5. tempname A 6. nwtomata `netname', mat(`A') 7. quietly generate double `generate' = . 8. mata: st_store((1::`nodes'), "`generate'", nw_reach2(`A')) 9. if "`network'" != "" { 10. mata: _diag(`A', 0) 11. mata: `A' = ((`A' + `A'*`A') :> 0) 12. mata: _diag(`A', 0) 13. local undirected = cond("`directed'" == "false", "undirected", "") 14. nwset, mat(`A') name(`network') labs(`labs') `undirected' 15. } 16. mata: mata drop `A' 17. quietly summarize `generate' 18. return scalar mean = r(mean) 19. return local netname "`netname'" 20. end . . drop reach2 . nwreach2 flomarriage, generate(reach2) network(floreach2b) . return list scalars: r(mean) = 6.875 macros: r(netname) : "flomarriage" . nwsummarize floreach2b -------------------------------------------------- Network name: floreach2b Network id: 5 Directed: false Valued: false Two-mode: false Nodes: 16 Selfloop: false Edges: 55 Minimum value: 0 Maximum value: 1 Density: .458 Mean degree: 6.88 Temporal: false . . * 5. Loop over the waves of a longitudinal network . nwwebuse glasgow, nwclear . matrix W = J(3, 1, .) . local i 0 . foreach w in glasgow1 glasgow2 glasgow3 { 2. local ++i 3. nwreach2 `w', generate(reach_`w') 4. matrix W[`i', 1] = r(mean) 5. } . matrix rownames W = glasgow1 glasgow2 glasgow3 . matrix colnames W = mean_reach2 . matrix list W W[3,1] mean_reach2 glasgow1 4.72 glasgow2 5 glasgow3 4.54 . summarize reach_* Variable | Obs Mean Std. dev. Min Max -------------+--------------------------------------------------------- reach_glas~1 | 50 4.72 3.162536 0 15 reach_glas~2 | 50 5 2.857143 0 13 reach_glas~3 | 50 4.54 3.018379 0 10 . . * 6. Pitfalls . * node order: the dataset is put back in node order by _nwdatasync . nwwebuse florentine, nwclear . sort wealth . list _nwnode wealth in 1/3 +---------------------+ | _nwnode wealth | |---------------------| 1. | pucci 3 | 2. | guadagni 8 | 3. | acciaiuoli 10 | +---------------------+ . _nwdatasync flomarriage . list _nwnode wealth in 1/3 +---------------------+ | _nwnode wealth | |---------------------| 1. | acciaiuoli 10 | 2. | albizzi 36 | 3. | barbadori 55 | +---------------------+ . . * directed: a row is a sender, a column a receiver . nwwebuse glasgow, nwclear . nwtomata glasgow1, mat(G) . mata: sum(G :!= G') 70 . _nwsyntax glasgow1 . display "directed = `directed', valued = `valued', nodes = `nodes'" directed = true, valued = false, nodes = 50 . . * valued: the matrix holds the values, not 0/1 . nwwebuse sampson, nwclear . nwsummarize samplike -------------------------------------------------- Network name: samplike Network id: 4 Directed: true Valued: true Two-mode: false Nodes: 18 Selfloop: false Arcs: 88 Minimum value: 0 Maximum value: 3 Density: .288 Mean outdegree: 4.89 Temporal: false . nwtomata samplike, mat(S) . mata: max(S) 3 . mata: sum(S :> 0) // 106, not 88 106 . mata: _diag(S, 0) . mata: sum(S :> 0) 88 . mata: mata drop G S . . * the diagonal . nwwebuse florentine, nwclear . nwtomata flomarriage, mat(A) . mata: (A*A)[1, 1..3] 1 2 3 +-------------+ 1 | . . . | +-------------+ . mata: _diag(A, 0) . mata: (A*A)[1, 1..3] 1 2 3 +-------------+ 1 | 1 1 1 | +-------------+ . . * memory . mata: mata describe # bytes type name and extent ------------------------------------------------------------------------------- 172 real colvector nw_nbrmean() 192 real colvector nw_reach2() 2,048 real matrix A[16,16] 8 class scalar __nw -------------------------------------------------------------------------------