@ Program to examine spatial correlation in New York City @ @ hate crime data. Analysis assumes a Gaussian model of @ @ the form Y=rY+XB+u, where XB=exp(x*b). Thus, the @ @ exogenous factors in the model are linked to hate crime @ @ via the same kind of exponential function assumed in @ @ the negative binomial analyses examined elsewhere. @ @ The models examined in this program look for spatial @ @ correlation net of the effects of demographic change @ open f1 = "temp"; d = readr(f1,rowsf(f1)); makevars(d,0,getname("temp")); f1 = close(f1); let n[51,51]; @ run greenopt (a Newton-Rapheson search algorithm) first @ #include greenopt; @-------------------------------------------------------------------@ @ create adjacency matrix @ @-------------------------------------------------------------------@ j=1; n[j,2]=1; n[j,3]=1; n[j,7]=1; n[j,33]=1; j=2; n[j,1]=1; n[j,3]=1; n[j,4]=1; n[j,5]=1 ; n[j,7]=1 ; j=3; n[j,1]=1; n[j,2]=1; n[j,4]=1; n[j,32]=1; n[j,34]=1; j=4; n[j,2]=1; n[j,3]=1; n[j,5]=1; n[j,34]=1; j=5; n[j,2]=1; n[j,4]=1; n[j,6]=1; n[j,10]=1; n[j,34]=1; j=6; n[j,5]=1; n[j,34]=1; j=7; n[j,1]=1; n[j,2]=1; n[j,8]=1; n[j,9]=1; j=8; n[j,7]=1; n[j,9]=1; j=9; n[j,7]=1; n[j,8]=1; n[j,10]=1; j=10;n[j,5]=1; n[j,9]=1; j=11;n[j,12]=1;n[j,13]=1;n[j,14]=1;n[j,36]=1; n[j,39]=1; j=12;n[j,11]=1;n[j,13]=1;n[j,16]=1;n[j,18]=1; j=13;n[j,11]=1;n[j,12]=1;n[j,14]=1;n[j,18]=1; n[j,26]=1; j=14;n[j,11]=1;n[j,13]=1;n[j,15]=1;n[j,26]=1; n[j,39]=1; j=15;n[j,14]=1;n[j,26]=1;n[j,28]=1;n[j,39]=1; n[j,43]=1;n[j,44]=1; j=16;n[j,12]=1;n[j,17]=1;n[j,18]=1; j=17;n[j,16]=1;n[j,20]=1;n[j,22]=1; j=18;n[j,12]=1;n[j,13]=1;n[j,16]=1;n[j,19]=1; n[j,26]=1; j=19;n[j,18]=1;n[j,27]=1; j=20;n[j,17]=1;n[j,21]=1;n[j,22]=1; j=21;n[j,20]=1;n[j,22]=1;n[j,23]=1;n[j,25]=1; j=22;n[j,17]=1;n[j,20]=1;n[j,21]=1;n[j,24]=1; j=23;n[j,21]=1;n[j,25]=1; j=24;n[j,22]=1;n[j,25]=1;n[j,27]=1;n[j,28]=1; j=25;n[j,21]=1;n[j,23]=1;n[j,24]=1;n[j,28]=1; j=26;n[j,13]=1;n[j,14]=1;n[j,15]=1;n[j,18]=1; n[j,27]=1;n[j,28]=1; j=27;n[j,19]=1;n[j,24]=1;n[j,26]=1;n[j,28]=1; j=28;n[j,15]=1;n[j,24]=1;n[j,25]=1;n[j,26]=1; n[j,27]=1; j=29;n[j,31]=1; j=30;n[j,33]=1; j=31;n[j,29]=1;n[j,32]=1;n[j,34]=1; j=32;n[j,3]=1 ;n[j,31]=1;n[j,33]=1; j=33;n[j,1]=1 ;n[j,30]=1;n[j,32]=1; j=34;n[j,3]=1 ;n[j,4]=1 ;n[j,5]=1 ;n[j,6]=1 ; n[j,31]=1; j=35;n[j,36]=1;n[j,37]=1; j=36;n[j,11]=1;n[j,35]=1;n[j,37]=1;n[j,38]=1; n[j,39]=1; j=37;n[j,35]=1;n[j,36]=1;n[j,38]=1; j=38;n[j,36]=1;n[j,37]=1;n[j,39]=1;n[j,40]=1; j=39;n[j,11]=1;n[j,14]=1;n[j,15]=1;n[j,36]=1;n[j,38]=1;n[j,40]=1; j=40;n[j,38]=1;n[j,39]=1;n[j,43]=1; j=41;n[j,42]=1;n[j,45]=1; j=42;n[j,41]=1;n[j,43]=1;n[j,45]=1;n[j,46]=1; j=43;n[j,15]=1;n[j,40]=1;n[j,42]=1;n[j,44]=1; n[j,46]=1; j=44;n[j,15]=1;n[j,43]=1;n[j,46]=1; j=45;n[j,41]=1;n[j,42]=1;n[j,47]=1; j=46;n[j,42]=1;n[j,43]=1;n[j,44]=1;n[j,47]=1; j=47;n[j,45]=1;n[j,46]=1; j=48; j=49;n[j,50]=1; j=50;n[j,49]=1;n[j,51]=1; j=51;n[j,50]=1; w=n; print " Analysis 1: Anti-Asian Hate Crime"; y=newahc; x=ones(rows(w),1)~pw80~chngasn~xasn; @-------------------------------------------------------------------@ @ set up likelihood function (see Doreian, Soc Meth) @ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in asian % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@ let p0={2.5,.02,0,.72,-17,42}; @ starting values @ proc qfct(p); @ specify the log-likelihood function @ local A,B,pred1; A=eye(rows(w))-p[2,1]*w; B=p[3:6,1]; pred1=exp(x*B); retp( - (rows(w)/2)*ln(p[1,1]^2) - (1/(2*p[1,1]^2))*(y'A'A*y - 2*pred1'A*y + pred1'pred1) + ln(det(A)) ); endp; { p,g,h,db,tol,f,pvcx } = greenopt(p0,&qfct); print" @-------------------------------------------------------------------@ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in asian % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@"; print " Analysis 2: Anti-Latino Hate Crime"; y=newlhc; x=ones(rows(w),1)~pw80~chnghsp~xhsp; @-------------------------------------------------------------------@ @ set up likelihood function (see Doreian, Soc Meth) @ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in latino % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@ let p0={3.4,.02,.3,1.3,-24,45}; @ starting values @ proc qfct(p); @ specify the log-likelihood function @ local A,B,pred1; A=eye(rows(w))-p[2,1]*w; B=p[3:6,1]; pred1=exp(x*B); retp( - (rows(w)/2)*ln(p[1,1]^2) - (1/(2*p[1,1]^2))*(y'A'A*y - 2*pred1'A*y + pred1'pred1) + ln(det(A)) ); endp; { p,g,h,db,tol,f,pvcx } = greenopt(p0,&qfct); print" @-------------------------------------------------------------------@ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in latino % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@"; print " Analysis 3: Anti-Black Hate Crime"; y=newbhc; x=ones(rows(w),1)~pw80~chngblk~xblk; @-------------------------------------------------------------------@ @ set up likelihood function (see Doreian, Soc Meth) @ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in black % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@ let p0={.15,.02,1.8,1.6,-4,16}; @ starting values @ proc qfct(p); @ specify the log-likelihood function @ local A,B,pred1; A=eye(rows(w))-p[2,1]*w; B=p[3:6,1]; pred1=exp(x*B); retp( - (rows(w)/2)*ln(p[1,1]^2) - (1/(2*p[1,1]^2))*(y'A'A*y - 2*pred1'A*y + pred1'pred1) + ln(det(A)) ); endp; { p,g,h,db,tol,f,pvcx } = greenopt(p0,&qfct); print" @-------------------------------------------------------------------@ @ p[1,1] == sigma @ @ p[2,1] == spatial autocorr parameter (rho) @ @ p[3,1] == intercept @ @ p[4,1] == main effect, % white @ @ p[5,1] == main effect, change in black % @ @ p[6,1] == interaction @ @-------------------------------------------------------------------@";