options linesize=100 pagesize=100;
libname revanch "c:\sasdat\revanch";
data revanch.citydat;
	infile "c:\sasdat\revanch\clusex.csv" delimiter=",";
	length fips $ 12;
	input fips $ city $ hchg78 hcgh89 hcgh90 pchg90 nhb9 hsp9 gq9 corr9 hless9 pov9	
    wpcapy9 wbratio9 own9 vac9 nhb0 hsp0 asn0 gq0 corr0 wdinks0 own0 ownchg90 
    wrent0 brent0 vac0 vacgh90;
run;
proc standard data=revanch.citydat mean=0 std=1 out=revanch.citystd;
    var hchg78 hcgh89 hcgh90 pchg90 nhb9 hsp9 gq9 corr9 hless9 pov9	
    wpcapy9 wbratio9 own9 vac9 nhb0 hsp0 asn0 gq0 corr0 wdinks0 own0 ownchg90 
    wrent0 brent0 vac0 vacgh90;
run;
proc factor data=revanch.citydat 
	method=principal nfactors=6 rotate=varimax out=revanch.scores;
	var hcgh90 pchg90 nhb9 hsp9 gq9 corr9 hless9 pov9	
    wpcapy9 wbratio9 own9 vac9 nhb0 hsp0 gq0 corr0 wdinks0 own0 
    wrent0 brent0 vac0;
	title "revised, rotated principal components analysis";
run;
proc cluster data=revanch.scores method=ward ccc pseudo print=30;
	var factor1 factor2 factor3 factor4 factor5 factor6;
	id fips;
	title "ward's analysis to determine number of clusters";
	run;
proc fastclus data=revanch.scores drift maxiter=500 maxclusters=12
	out=revanch.trctclus;
	var factor1 factor2 factor3 factor4 factor5 factor6;
	id fips;
	title "final tract classifications, 12 clusters";
	run;
proc contents data=revanch.trctclus;
run;
proc freq data=revanch.trctclus;
	tables cluster*city;
	title "gentrification type by city";
run;
proc sort data=revanch.trctclus;
	by cluster distance;
run;
proc print data=revanch.trctclus;
	var cluster distance fips city hcgh90 pchg90 nhb9 hsp9 gq9 corr9 hless9 pov9	
    wpcapy9 wbratio9 own9 vac9 nhb0 hsp0 gq0 corr0 wdinks0 own0 
    wrent0 brent0 vac0;
run;
proc means data=revanch.trctclus;
	var hcgh90 pchg90 nhb9 hsp9 gq9 corr9 hless9 pov9	
    wpcapy9 wbratio9 own9 vac9 nhb0 hsp0 gq0 corr0 wdinks0 own0 
    wrent0 brent0 vac0;
by cluster;
run;

data revanch.visclus;
set revanch.trctclus;
length color $ 15;

if cluster=1 then color="red";
if cluster=3 then color="blue";
if cluster=5 then color="yellow";
if cluster=9 then color="green";
if cluster=12 then color="magenta";

if cluster=2 then color="white";
if cluster=4 then color="white";
if cluster=6 then color="white";
if cluster=7 then color="white";
if cluster=8 then color="white";
if cluster=10 then color="white";
if cluster=11 then color="white";
run;

goptions cback=black colors=(white white magenta yellow green);
title1 "Classification of Gentrified Tracts";

proc g3d data=revanch.visclus;
	scatter factor1*factor2=factor3/
	color=color;
	label factor1="segregation" factor2="redevelopment"
		  factor3="tenure";
	footnote color=red "cluster 1 "
		 	 color=blue "cluster 3 "
		 	 color=yellow "cluster 5 "
		 	 color=green "cluster 9 "
			 color=magenta "cluster 12 "
			 color=white "outliers "
		 ;
	run;
proc g3d data=revanch.visclus;
	scatter factor2*factor3=factor4/
	color=color;
	label factor2="redevelopment" factor3="tenure"
		  factor4="institutions";
	footnote color=red "cluster 1 "
		 	 color=blue "cluster 3 "
		 	 color=yellow "cluster 5 "
		 	 color=green "cluster 9 "
			 color=magenta "cluster 12 "
			 color=white "outliers "
		 ;
	run;
proc g3d data=revanch.visclus;
	scatter factor3*factor4=factor5/
	color=color;
	label factor3="tenure" factor4="institutions"
		  factor5="latinos";
	footnote color=red "cluster 1 "
		 	 color=blue "cluster 3 "
		 	 color=yellow "cluster 5 "
		 	 color=green "cluster 9 "
			 color=magenta "cluster 12 "
			 color=white "outliers "
		 ;
	run;
proc g3d data=revanch.visclus;
	scatter factor4*factor5=factor6/
	color=color;
	label factor4="institutions" factor5="latinos"
		  factor6="inequality";
	footnote color=red "cluster 1 "
		 	 color=blue "cluster 3 "
		 	 color=yellow "cluster 5 "
		 	 color=green "cluster 9 "
			 color=magenta "cluster 12 "
			 color=white "outliers "
		 ;
	run;