Spatial Regression Models, Version 2 Chapter 3 Michael D. Ward and Kristian Skrede Gleditsch

Size: px

Start display at page:

Download "Spatial Regression Models, Version 2 Chapter 3 Michael D. Ward and Kristian Skrede Gleditsch"

Barrie Carson
5 years ago
Views:

1 Spatial Regression Models, Version 2 Chapter 3 Michael D. Ward and Kristian Skrede Gleditsch Introduction These are lecture/lab notes for Chapter 3 in Spatial Regression Models. You will find all necessary information, including R-code, to replicate the figures and maps found in this chapter. A few of the maps were obtained from public repositories (Wiki and others), but what you find here will allow replication of the bulk of chapter 3. This is intended to provide code as well as a didactic repository. Code appears below as it is in the unadorned code files. Comments are shown with one poundsign; messages from the execution of this code is flagged with two poundsigns. There are two goals of this material: 1. Provide replication code for the materials in Spatial Regression Models, version 2. This is hosted on httpps://srmbook.com 2. To elaborate a bit on what the code does to provide instruction on develop code in R that replications the material in the published volume. This was developed on the following platform. Earlier (and later) platforms should work (as should windows and linux flavored platforms). However, some libraries may not be available for all platforms. Such is life. If you have any feature requests or find bugs herein, please do not hesitate to pass them along to the authors (michael.don.ward at gmail.com and ksg at essex.ac.uk). Figure 3.1 Polity Scores as a Function of Logged GDP per Capita A longstanding hypothesis in sociology and political science is that economic development leads to the establishment of democratic institutions. If we make an economic assumption that we observe the world in equilibrium this implies that countries with higher levels of economic productivity will have higher levels of democratic governance. We re-examine this by looking at GDP per capita and the level of democracy in 195 contemporary (2015) countries.p5 rm(list = ls()) library(foreign) library(ggplot2) library(gridextra) library(grid) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) # get map from CSHAPES 1

2 Figure Polity Score GDP, logged per capita Figure 1: Polity Scores as a Function of Logged GDP per Capita wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame # Table 3.1 wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc dat<-subset(wmap.df,select = c("polity","gdp")) #ratio.display <- 4/3 #ratio.values <- ( )/( ) +.1 # open png device ggplot(dat, aes(x=gdp, y=polity)) + geom_point(shape=16) + geom_smooth(method=lm) + labs(x = "GDP, logged per capita",y = "Polity Score") + coord_fixed(.2) + ggtitle("figure 3.1") 2

3 Figure 3.2, Examination of the Normality of the Residuals opts_knit$set(root.dir = "~/Dropbox/SRMRepository") library(foreign) library(ggplot2) library(gridextra) library(grid) library(envstats) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame # Table 3.1 wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc dat<-subset(wmap.df,select = c("polity","gdp")) ols.1.2<-glm(polity ~gdp,data=dat) qqnorm(resid(ols.1.2),pch=19,las = 1,cex=.7,plot.type="Q-Q",points.col="black", ylab="sample Quantiles",xlab="Theoretical Quantiles", main="figure 3.2") segments(-2.9, -15, x1 = 1, y1 = 8, col = par("fg"), lty = par("lty"), lwd = par("lwd")) # Figure 3.3 Figure 3.3, Examination of the Normality of the Residuals, redux 3

4 Figure Sample Quantiles Theoretical Quantiles Figure 2: Examination of the Normality of the Residuals 4

5 Figure Figure 3: Examination of the Normality of the Residuals, redux plot(density(residuals(ols.1.2)),ylim=c(0,.5), xlim=c(-20,10),bty="n",axes=f,xlab="",ylab="", lwd=3,main="figure 3.3") lines(density(rnorm(100000,0,1)),lwd=3,col="gray") axis(side=1,at=c(-20,-10,0,10)) Figure 3.4, A Simple Network Visualization of the Linkages among Eight European Countries library(ggally) library(network) 5

6 UK FRN GER ITA SWE FIN DEN NOR Figure 4: A Simple Network Visualization of the Linkages among Eight European Countries library(sna) library('igraph') library(foreign) library(ggplot2) cshp.data <- cshp() easyeight.shp <- cshp.data[cshp.data$cowcode%in% c(200,220,260,325,375,380,385,390),] wmat.195 <- distmatrix(as.date(" "), type="mindist", tolerance=0.5, usegw=true) ee<-as.character(c(200,220,260,325,375,380,385,390)) wmat.8 <- wmat.195[ee,ee] w8.dist<-wmat.8 w8.dist[w8.dist<=200]<-1 w8.dist[w8.dist>200]<-0 colnames(w8.dist)<-row.names(w8.dist)<-c("uk","frn","ger","ita","fin","swe","nor","den") diag(w8.dist)<-0 set.seed(12345) net.igraph <- graph_from_adjacency_matrix(w8.dist, plot(net.igraph,vertex.label.cex=.6, vertex.color = "skyblue") mode = c("undirected"), diag = FALSE) Figure 3.5, A Simple Network Visualization of the Linkags among 195 Countries 6

7 library(ggally) library(network) library(sna) library('igraph') library(foreign) library(ggplot2) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc wmat.195 <- distmatrix(as.date(" "), type="mindist", tolerance=0.5, usegw=true) wmat.195[wmat.195<=200]<-1 wmat.195[wmat.195>200]<-0 diag(wmat.195)<-0 wmap.df.sorted<-wmap.df[order(wmap.df$gwcode),] colnames(wmat.195)<-row.names(wmat.195)<-wmap.df.sorted$iso1al3 net195<-network(wmat.195) net.igraph <- graph_from_adjacency_matrix(wmat.195, mode = c("undirected"), diag = FALSE) sizes<-wmap.df$gdp sizes[is.na(sizes)]<-5.75 V(net.igraph)$gdp=sizes*1.5 V(net.igraph)$size <- degree(net.igraph) ##V(net.igraph)$gdp iso <- V(net.igraph)[degree(net.igraph)==0] g2 <- delete.vertices(net.igraph, iso) plot(g2,vertex.label.cex=.6, vertex.color = "skyblue") 7

8 DMA LCA SURURY BRB VCTGRDTTO GUY VEN BRA ARG PRY LSO BOL CHL COL PER SWZZAF NAM ECU ZWE BWA PANCRI ATG MOZ GNQ GHA LBRSLE MWI ZMB AGO NIC KNA GAB TGO BENBFA CIV GIN GNB COG TZACOD CMRNGA GMB HND SLV RWA BDI GTM BLZ DOM CAF NER MLI SEN UGA SSD TCD HTI KEN LKA MEX MRT CUB ETH SDN JAM SOMERI LBY DZA IND BTN NPL CAN DJI EGY BGD YEM MAR JOR KWT PAK USA OMN ISRSAU IRQ TKM AFG TJK MMR KGZ LAO UZB CHN VNM SYR IRN THA KHM PHL ARE LBN KAZ TUN ARM CYP GEO BHR QAT TUR ESPAZE MNG TWN MYSBRN PRT RUS PRK SGP IDNTLS GRC BGRMDA EST JPN UKR AND BLR LTU LVA KOR PNG RKS MKD ALB SRB ROU POL FIN MNE NOR BIHITASVK HUN HRV CZE FRA SWE MLT SVN AUT DNK LIE CHE DEU SMR MCO BEL LUX NLD GBR IRL Figure 5: A Simple Network Visualization of the Linkags among 195 Countries Figure 3.6 Map of Centroids and Capital Cities library(rcolorbrewer) library(maptools) library(spdep) library(sp) library(rgdal) # Read a Robinson projection map from an ESRI shapefile newprojection <- "+proj=robin +lon_0=0 +x_0=0 +y_0=0" #rob.shp <- readogr(dsn="shapefiles/110m_cultural", layer="ne_110m_admin_0_countries") rob.shp <- readshapespatial("shapefiles/world2002/wg2002worldmap.shp",proj4string = CRS(newprojection), IDvar = "FIPS_CNTRY",repair=TRUE,force_ring=TRUE, verbose=true) ## Shapefile type: Polygon, (5), # of Shapes: 190 ## Shapefile type: Polygon, (5), # of Shapes: 190 # Add coordinates coords <- coordinates(rob.shp) # Replot the map itself without a bounding box plot(rob.shp, border="grey", xaxt="n", yaxt="n", bty="n", lwd=.75, las=1, ylab="", 8

9 Centroids and Capitals Figure 6: Map of Centroids and Capital Cities main="centroids and Capitals", xlab="") # Add the centroids points(coordinates(rob.shp), pch=19,cex=.25, col="red") # Latitude and Longitude ll.mat <- cbind(as.numeric(as.character(rob.shp$long)), as.numeric(as.character(rob.shp$lat))) row.names(ll.mat) <- 1:nrow(ll.mat) llcrs <- CRS("+proj=longlat +ellps+wgs84") longlat <- SpatialPoints(ll.mat,proj4string = llcrs) longlat.transf <- sptransform(longlat, CRS("+proj=robin +ellps+wgs84+lon_0=0 + x_0=0 +y_0=0")) points(coordinates(longlat.transf),pch=19,cex=.25,col="black") # Add segments between centroids and capitals # but delete Kiribati, as it crosses date line diffs <- cbind(coordinates(rob.shp),coordinates(longlat.transf)) segments(diffs[-91,1],diffs[-91,2],diffs[-91,3],diffs[-91,4],col="slategray4") Figure 3.7 Anselin-Moran Plot of OLS Residuals 9

10 library(foreign) library(ggplot2) library(gridextra) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc # Prune Data Set moran.df<-subset(wmap.df,select = c("gwcode","polity","gdp")) pruned<-na.omit(moran.df) pruned<-pruned[with(pruned, order(gwcode)), ] row.names(pruned)<-pruned$gwcode ols.1.2<-glm(polity ~ gdp,data=pruned) wmap.dist <- distmatrix(date = as.date(" "),type="mindist") index<-as.character(pruned$gwcode) w.dist<-wmap.dist[index,index] # recode distance data to be binary, for distance < 550km w.dist[w.dist<=400]<-1 w.dist[w.dist>=400]<-0 w.dist.rowstd<-w.dist/rowsums(w.dist) # create spatial lag of residuals resid.splag<- w.dist.rowstd%*%resid(ols.1.2) resid.splag.std<-scale(resid.splag) resid.std<-scale(resid(ols.1.2)) plot(resid.std,resid.splag.std,las=1,pch=19,xlim=c(-3,3), ylim=c(-3,3),ylab="standardized Spatial Lag of OLS Residuals", xlab="standardized OLS Residuals",main="Moran's I = 0.72") polygon(c(0,0,1,1),c(0,1,1,0),col=rgb( 135/255,206/255,250/255,0.5)) polygon(c(-1,-1,0,0),c(-1,0,0,-1),col=rgb( 135/255,206/255,250/255,0.5)) points(resid.std,resid.splag.std,pch=19) segments(-2, -2, -2, 2,lwd=2) segments(2, -2, 2, 2,lwd=2) segments(-2, 2, 2,2,lwd=2) segments(-2, -2, 2,-2,lwd=2) 10

11 Moran's I = 0.72 Standardized Spatial Lag of OLS Residuals Standardized OLS Residuals Figure 7: Anselin-Moran Plot of OLS Residuals segments(-1, -1, -1, 1,lwd=2) segments(1, -1, 1, 1,lwd=2) segments(-1, 1, 1,1,lwd=2) segments(-1, -1, 1,-1,lwd=2) segments(-2, 0, 2,0,lwd=2) segments(0, -2,0,2,lwd=2) rug(resid.std, ticksize = 0.03, side = 1, lwd = 0.5) rug(resid.splag.std, ticksize = 0.03, side = 2, lwd = 0.5) abline(lm(resid.splag.std~-1+resid.std),lwd=2) Figure 3.8 Choropleth of Local Indicators of Spatial Autocorrelation 11

12 [ 10, 6) [ 6, 3) [ 3,0) [0,3) [3,6) [6,10] Figure 8: Choropleth of LISA Scores p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap@data$democ <- p5$democ[match(wmap@data$iso1al3, p5$scode)] wmap@data$autoc <- p5$autoc[match(wmap@data$iso1al3, p5$scode)] wmap@data$polity <- wmap@data$democ - wmap@data$autoc # for convenience assign 0 to NA wmap@data$polity[is.na(wmap@data$polity)]<- 0 pol <- wmap@data$polity library(classint) nclr <- 6 plotclr <- brewer.pal(nclr,"spectral") class = classintervals(pol, n = 6, style = "fixed", fixedbreaks = c(-10, -6, -3, 0, 3, 6, 10)) colcode <- findcolours(class, plotclr) plot(wmap, col=colcode) legend(-170,0, legend=names(attr(colcode, "table")), fill=attr(colcode, "palette"), cex=0.6, bty="n") # now get distance matrix wmap.dist <- distmatrix(date = as.date(" ")) # recode distance data to be binary, for distance < 400km w.dist<-wmap.dist w.dist[w.dist<=1900]<-1 w.dist[w.dist>1900]<-0 # row standardize w.dist.rstd<-w.dist/rowsums(w.dist) # 0 on diagonal diag(w.dist.rstd)<-0 which(rowsums(w.dist.rstd)==0) ## named integer(0) w.dist.rstd <- w.dist.rstd[as.character(wmap@data$gwcode),as.character(wmap@data$gwcode)] # create spatial lag of polity. round result wmap@data$sp.lag.polity<- round((w.dist.rstd)%*%(wmap@data$polity)) 12

13 [ 1.4,0) [0,0.18) [0.18,0.549) [0.549,0.85) [0.85,3.72] Figure 9: Choropleth of LISA Scores lmoran <- mat2listw(w.dist.rstd)) pol <- library(classint) nclr <- 6 plotclr <- brewer.pal(nclr,"prgn") class = classintervals(pol, n = 6, style = "fixed", fixedbreaks = c(-1.4,0.0, 0.18,0.549,0.85,3.72 )) colcode <- findcolours(class, plotclr) plot(wmap, col=colcode) legend(-170,0, legend=names(attr(colcode, "table")), fill=attr(colcode, "palette"), cex=0.6, bty="n") Table 3.1 top 10 on polity and gdp in terms of Polity library(foreign) library(ggplot2) library(gridextra) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) 13

14 gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame # Table 3.1 wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc # Prune Data Set moran.df<-subset(wmap.df,select = c("gwcode","isoname","polity","gdp")) pruned<-na.omit(moran.df) pruned<-pruned[with(pruned, order(gwcode)), ] row.names(pruned)<-pruned$gwcode ols.1.2<-glm(polity ~ gdp,data=pruned) sorted<-pruned[order(pruned$gdp),] sorted$gdp<-round(sorted$gdp,2) bottom<-head(sorted,10) bottom<-bottom[,c(2:4)] levels(bottom$isoname)[40] <- 'Congo (DRC)' bottom[order(bottom$gdp,decreasing =TRUE),] ISONAME polity gdp 530 Ethiopia Liberia Congo (DRC) Madagascar Gambia Niger Somalia Central African Republic Malawi Burundi top<-tail(sorted,10) top<-top[,c(2:4)] top<-top<-top[order(bottom$gdp,decreasing =TRUE),] library(xtable) tab<-xtable(rbind(top,bottom),caption="democracy Data and Logged GDP per capita. Top and Bottom Ten Countri print(tab,caption.placement="top") % latex table generated in R by xtable package % Wed Jan 17 14:44: Table 3.2 library(foreign) library(ggplot2) library(gridextra) library(grid) library(envstats) library(xtable) 14

15 Table 1: Democracy Data and Logged GDP per capita. Top and Bottom Ten Countries ISONAME polity gdp 212 Luxembourg Qatar Norway Switzerland Denmark Australia Sweden Singapore Ireland United States Burundi Malawi Central African Republic Somalia Niger Gambia Madagascar Liberia Congo (DRC) Ethiopia p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) wmap@data$id <- as.character(rownames(wmap@data)) wmap.df <- wmap@data wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] library(wbstats) gdp <- wb(indicator = c("ny.gdp.pcap.cd"), startdate = 2014, enddate = 2014,country="countries_only") gdp$value <- log(gdp$value) # join the GDP data to the map data.frame # Table 3.1 wmap.df$gdp <- gdp$value[match(wmap.df$iso1al2, gdp$iso2c)] # create polity score wmap.df$polity<-wmap.df$democ - wmap.df$autoc dat<-subset(wmap.df,select = c("polity","gdp")) ols.1.2<-glm(polity ~gdp,data=dat) tab<-xtable(ols.1.2,digits=2,caption="ols Regression of Polity on Logged GDP per capita.") print(tab,caption.placement="top") % latex table generated in R by xtable package % Wed Jan 17 14:17:

16 Table 2: OLS Regression of Polity on Logged GDP per capita. Estimate Std. Error t value Pr(> t ) (Intercept) gdp Table 3.3, A List Representation of Connections for Eight European Countries library(network) library(sna) cshp.data <- cshp() easyeight.shp <- cshp.data[cshp.data$cowcode%in% c(200,220,260,325,375,380,385,390),] wmat.195 <- distmatrix(as.date(" "), type="mindist", tolerance=0.5, usegw=true) ee<-as.character(c(200,220,260,325,375,380,385,390)) wmat.8 <- wmat.195[ee,ee] w8.dist<-wmat.8 w8.dist[w8.dist<=200]<-1 w8.dist[w8.dist>200]<-0 colnames(w8.dist)<-row.names(w8.dist)<-c("uk","frn","ger","ita","fin","swe","nor","den") diag(w8.dist)<-0 tmp<-mat2listw(w8.dist) tab<-xtable(w8.dist) print(tab,caption.placement="top") % latex table generated in R by xtable package % Wed Jan 17 14:17: UK FRN GER ITA FIN SWE NOR DEN UK FRN GER ITA FIN SWE NOR DEN Table 3.4, (Adjacency) Matric Representation of Connections for Eight European Countries 16

17 library(network) library(sna) cshp.data <- cshp() easyeight.shp <- cshp.data[cshp.data$cowcode%in% c(200,220,260,325,375,380,385,390),] wmat.195 <- distmatrix(as.date(" "), type="mindist", tolerance=0.5, usegw=true) ee<-as.character(c(200,220,260,325,375,380,385,390)) wmat.8 <- wmat.195[ee,ee] w8.dist<-wmat.8 w8.dist[w8.dist<=200]<-1 w8.dist[w8.dist>200]<-0 colnames(w8.dist)<-row.names(w8.dist)<-c("uk","frn","ger","ita","fin","swe","nor","den") diag(w8.dist)<-0 tmp<-matrix(as.integer(w8.dist),nrow=8) colnames(tmp)<-row.names(tmp)<-c("uk","frn","ger","ita","fin","swe","nor","den") tab<-xtable(tmp,caption="(adjacency) Matrix Representation of Connections for 8 European Countries.") print(tab,caption.placement="top") % latex table generated in R by xtable package % Wed Jan 17 14:18: Table 3: (Adjacency) Matrix Representation of Connections for 8 European Countries. UK FRN GER ITA FIN SWE NOR DEN UK FRN GER ITA FIN SWE NOR DEN Table 3.6 Democracy Data (PITF: 2015) library(foreign) library(ggplot2) library(gridextra) p5<-read.csv("rawdata/p5.csv") wmap <- cshp(date = as.date(" "),usegw=true) # get map from CSHAPES 17

18 wmap$iso1al3 <- as.character(wmap$iso1al3) wmap$iso1al3[wmap$cntry_name == "Kosovo"] <- c("rks") wmap$iso1al3 <- as.factor(wmap$iso1al3) <- wmap.df <- wmap.df$democ <- p5$democ[match(wmap.df$iso1al3, p5$scode)] wmap.df$autoc <- p5$autoc[match(wmap.df$iso1al3, p5$scode)] wmap.df$polity<-wmap.df$democ - wmap.df$autoc # Prune Data Set temp<-subset(wmap.df,select = c("gwcode","isoname","polity")) pruned<-na.omit(temp) pruned<-pruned[with(pruned, order(gwcode)), ] row.names(pruned)<-pruned$gwcode # wdist has the distance scores, in order of GW codes wmap.dist <- distmatrix(date = as.date(" "),type="mindist") index<-as.character(pruned$gwcode) w.dist<-wmap.dist[index,index] # recode distance data to be binary, for distance < 550km w.dist[w.dist<=400]<-1 w.dist[w.dist>=400]<-0 w.dist.rowstd<-w.dist/rowsums(w.dist) # create spatial lag of residuals pruned$polity.splag<- round(w.dist.rowstd%*%pruned$polity,2) # Prune Data Set tab.df<-subset(pruned,select = c("gwcode","isoname","polity","polity.splag")) pruned<-pruned[with(pruned, order(gwcode)), ] row.names(pruned)<-pruned$gwcode # now calculate the spatial lags, and add back to pruned. sorted<-pruned[order(pruned$polity.splag),] bottom<-head(sorted,10) bottom<-bottom[,c(2:4)] levels(bottom$isoname)[98] <- 'Laos' bottom<-bottom[order(bottom$polity.splag,decreasing =TRUE),] top<-tail(sorted,10) top<-top[,c(2:4)] top<-top[order(top$polity.splag,decreasing =TRUE),] tab<-xtable(rbind(top,bottom),caption="democracy data (PITF: 2015). Top and bottom 10 countries in terms of spatially lagged Polity scores.") print(tab,caption.placement="top") % latex table generated in R by xtable package % Wed Jan 17 15:02:

19 Table 4: Democracy data (PITF: 2015). Top and bottom 10 countries in terms of spatially lagged Polity scores. ISONAME polity polity.splag 20 Canada Ireland Cape Verde Mauritius New Zealand Belgium Luxembourg France United Kingdom Netherlands Yemen Saudi Arabia Cambodia Laos Sudan Oman Kuwait Bahrain Qatar United Arab Emirates

Inferring Latent Preferences from Network Data

Inferring Latent Preferences from Network John S. Ahlquist 1 Arturas 2 1 UC San Diego GPS 2 NYU 14 November 2015 very early stages Methodological extend latent space models (Hoff et al 2002) to partial