# edps 587
# Spring 2019
# c.j.anderson
#
# Introduction:  JSP data

setwd("D:\\Dropbox\\edps587\\\\Lectures\\1 introduction")

jsp <- read.csv("jsp_data.csv",header=TRUE,sep=",")
jsp <- na.omit(jsp)


# Pull out data for school 1, years 0 and 1
jsp1 <- jsp[ which(jsp$school==1 & jsp$jsyear!=2), ]

# change long to wide
jsp.wide1 <- reshape(jsp1, idvar = "id", timevar = "jsyear", direction = "wide")

# plot points for school 1
plot(jsp.wide1$math.0,jsp.wide1$math.1, type="p", pch=19,
     main=expression(paste("JSP Data,      (", R^2, "= .70)")),
	 xlab="Math Scores Year=0",
	 ylab="Math Scores Year=1",
	 xlim=c(0,50),
	 ylim=c(0,50)
	 )
abline(lm(jsp.wide1$math.1 ~ jsp.wide1$math.0),col="blue",lwd=2)
summary(lm1<- lm(jsp.wide1$math.1 ~ jsp.wide1$math.0))
text(19,47.5, expression(paste(math[1]," = 3.4251 + 0.8803*",math[0])),cex=1.25)



##########################################
# Graph regression lines all in one plot #
##########################################

# Pull out data for years 0 and 1  (all school)
jsp.2 <- jsp[ which(jsp$jsyear!=2), ]

# Change from long to wide
jsp.all <- reshape(jsp.2, idvar = "id", timevar = "jsyear", direction = "wide")
jsp.all <- na.omit(jsp.all)

school <- unique(jsp.all$school.0)                        
N <- length(school)                                  


# Graph to illustrate need for random intercept and slope
# -- create frame for plot, notice that type='n' for none 
plot(jsp.all$math.0,jsp.all$math.1,
	type="n", 
	col="blue",
	lwd=2,
	main="JSP: Linear Regression by School",
	xlab="Math Scores in Year = 0",
	ylab="Math Scores in Year = 1"
	)

# -- loop through the N schools and plot regression line for each
for (j in 1:N) {
  sub <- subset(jsp.all,jsp.all$school.0==j)
  fitted <- fitted(lm(math.1~math.0,data=sub))
  lines(sub$math.0,fitted,col=j)
}