guns.R

#SAMPLE CODE FOR SPATIAL PATTERNS OF VARIABLES

library(ggplot2)
library(fiftystater)
bachdeg <- read.csv("C:/Users/DASA0/Desktop/Stat 524/Project/bachdeg.csv", sep=",")
bachdeg$state <- tolower(bachdeg$State)
# map_id creates the aesthetic mapping to the state name column in your data
p <- ggplot(bachdeg, aes(map_id = state)) + 
  # map points to the fifty_states shape data
  geom_map(aes(fill = bachdeg), map = fifty_states) + 
  expand_limits(x = fifty_states$long, y = fifty_states$lat) +
  coord_map() + scale_fill_gradient(low='yellow', high='red') +
  scale_x_continuous(breaks = NULL) + 
  scale_y_continuous(breaks = NULL) +
  labs(x = "", y = "") +
  theme(legend.position = "bottom", 
        panel.background = element_blank())
P1 <- p + theme_bw()  + labs(fill = "Bachelor's degree" 
                             ,title = "People with at least a Bachelor's degree per 100,000 population, 2014", x="", y="")

# P.S.: OTHER VARIABLES CAN BE SIMILARLY PLOTTED USING SAME CODE

#SAMPLE CODE FOR CLUSTER ANALYSIS

data2 <- read.csv("C:/Users/DASA0/Desktop/Stat 524/Project/data2.csv", sep=",")
data2 <- as.data.frame(data2)
library(cluster)
data_new3 <- cbind( scale(data2[,2:9], FALSE, apply(data2[,2:9], 2, sd)), data2[10])
distmat3 <- dist(data_new3, method="euclidean")
fit.comp1 <- hclust(distmat3, method="average")
plot(fit.comp1, labels=data2$stateName)
member.fitcomp1 <- cutree(fit.comp1, 4)
aggregate(data_new3, list(member.fitcomp1), mean)
round(aggregate(data2[,-c(1,1)], list(member.fitcomp1), mean),2)
plot(silhouette(cutree(fit.comp1,4),distmat3))

fit.comp2 <- hclust(distmat3, method ="mcquitty")
plot(fit.comp2, labels=data2$stateName)
member.fitcomp2 <- cutree(fit.comp2, 4)
aggregate(data_new3, list(member.fitcomp2), mean)
round(aggregate(data2[,-c(1,1)], list(member.fitcomp2), mean),2)
plot(silhouette(cutree(fit.comp2,4),distmat3))

fit.comp3 <- hclust(distmat3, method="complete")
plot(fit.comp3, labels=data2$stateName)
member.fitcomp3 <- cutree(fit.comp3, 4)
aggregate(data_new3, list(member.fitcomp3), mean)
round(aggregate(data2[,-c(1,1)], list(member.fitcomp3), mean),2)
plot(silhouette(cutree(fit.comp2,4),distmat3))

fit.comp4 <- hclust(distmat3, method ="ward.D")
plot(fit.comp4, labels=data2$stateName)
member.fitcomp4 <- cutree(fit.comp4, 4)
aggregate(data_new3, list(member.fitcomp4), mean)
round(aggregate(data2[,-c(1,1)], list(member.fitcomp4), mean),2)
plot(silhouette(cutree(fit.comp4,4),distmat3))

fit.comp5 <- hclust(distmat3, method ="ward.D2")
plot(fit.comp5, labels=data2$stateName)
member.fitcomp5a <- cutree(fit.comp5, 4)
aggregate(data_new3, list(member.fitcomp5a), mean)
round(aggregate(data2[,-c(1,1)], list(member.fitcomp5a), mean),2)
plot(silhouette(cutree(fit.comp5,4),distmat3))
member.fitcomp5b <- cutree(fit.comp5, 5)
aggregate(data_new3, list(member.fitcomp5b), mean)
plot(silhouette(cutree(fit.comp5,5),distmat3))



#Scree Plot
wss <- (nrow(data_new3)-1)*sum(apply(data_new3, 2, var))
for (i in 2:48) wss[i] <- sum(kmeans(data_new3, centers=i)$withinss)
p <- plot(1:48, wss, type="b", main = "Scree Plot", xlab="Number of Clusters", ylab = "Within Group SS", pch=16)

#Correlation among variables
crime_mat <- data.matrix(data_new3[,1:8])
simmat <- round(cor(crime_mat), 3)
lower.tri(simmat, diag=FALSE)
upper<-simmat
upper[upper.tri(simmat)]<-""
upper<-as.data.frame(upper)
distmat <- as.dist(1-simmat)
fit.comp <- hclust(distmat, method="complete")
plot(fit.comp)

#Scatter Plot
pairs(~ violcr+ propcr+ medinc + ymunemp+ gini + bachdeg,data=data2)

This is the R code for an Exploratory analysis project I worked on. Write up for the project is here (with some changes compared to the project I submitted at school)- https://nashstatistica.wordpress.com/2018/05/27/how-i-stopped-worrying-bought-a-gun/