#----------------------------------------------------------------------
# Baseball examples for R and RStudio demo
#----------------------------------------------------------------------


## Useful option (if working in the command line on a wide terminal)
## This will change the number of columns before R wraps the line.
## NOTE: This & some other code below might contain items surrounded
##       by angle brackets (<...>). These should be replaced with actual
##       values before running.
#option(width=<# of columns>)


## Set working directory (uncomment if needed)
#setwd("<path to your working directory>")


## Read data into R
Data = read.table("hw1/RunsPerGame.txt", header=T)
str(Data)          # Get info about column names & types
summary(Data)      # Summary stats for each column

### ACCESSING DATA
## Rows and columns can be accessed by number...
print( Data[1,] )
print( Data[,1] )
print( Data[1,1:6] )

## Columns can be accessed by name
print( Data$League )
print( Data$League[1:10] )
print( Data[, 'League'] )
print( Data[,c('Team','League')] )

## Can combine to slice data as you'd like
print( Data[c(1:10,15:20), c('Team','League','AVG')] )

## Can select using conditionals
print( Data[Data$League=='American',] )
print( subset(Data, League == 'American') )
summary(Data$RPG[Data$League=='National'])

### DATA MANIPULATION
## Order the Data
Data <- Data[order(Data$Team),]

### COMMON CONTROL STRUCTURES
for ( team in Data$Team ) {
  print( team )
}
for ( i in 1:nrow(Data) ) {
  if ( Data$League[i] == "American") {
    print( paste( Data$Team[i],":", Data$RPG[i] ) )
  } 
  else {
    print( paste("SENIOR CIRCUIT", ":", Data$Team[i] ) )
  }
}

## Create R Function
Calculate.RunsPerAB <- function(x) {
  return(x$R/x$AB)
}
print( Calculate.RunsPerAB(Data) )

## Use attach() to avoid having to explicitly access by column name
##
## WARNING: Read the docs about "read-only" behavior. Also, attach()
## can clutter your namespace/environment (may create problem if
## you're dealing with multiple data sets)
attach(Data)
summary(RPG)
summary(RPG[League=='National'])


## Basic statistics
mean(RPG)
median(RPG)
sd(RPG)
var(RPG)
cov(OBP, RPG)
IQR(RPG)   # interquartile range -> useful for outlier detection


#### BASIC PLOTTING
## Basic plotting
boxplot(RPG)

## Multiple plots on the same "page"
par(mfrow=c(1,2))
boxplot(RPG, main='Runs per game')
boxplot(OBP, main='On base percentage')


## More complicated boxplots
par(mfrow=c(1,1))  # back to only one plot per page
boxplot(RPG ~ League)


## Scatter plot
## TIP: Read the docs for "cex" and "pch" options (change point types)
plot(AVG, RPG, cex=0.5, pch=19)


## Scatter plots with multiple variables
## matplot(...) might be useful in other cases
pairs(cbind(AVG,SLG,OBP,RPG), pch=19, cex=0.5)

## Basic linear regression
modelH = lm(RPG~H)

## Get info about regression model & plot
summary(modelH)
plot(H, RPG, pch=19, cex=1.2, xlab="Hits", ylab = "Runs per Game")
abline(modelH$coef[1], modelH$coef[2], col=2, lwd=2)


## Model evaluation
predict(modelH)    # Predicted values w/ training data
sqrt(mean(residuals(modelH)^2))  # Training RMSE
summary(modelH)$sigma # Residual Standard Error - uses degrees of freedom

# Q: How do we calculate RMSE for testing data?
# Hint: Can't use residuals(), but predict() does have a "newdata" param.

## Multi-parameter (predictor) models
model.H.SO = lm(RPG ~ H + SO)
summary(model.H.SO)

## Getting help on function parameters
?lm
# -- or --
help(lm)

## More principled way using stepwise regression
help.search('stepwise')
# -- or --
?? 'stepwise'

## GGPLOT - Slightly more elegant plotting
require(ggplot2)

## Histogram
ggplot(Data, aes(x=RPG)) + geom_histogram(binwidth=0.25)
## Different ways to view the effect of the League
ggplot(Data, aes(x=RPG, fill=League)) + geom_histogram(binwidth=0.25)
ggplot(Data, aes(x=RPG, fill=League)) + 
  geom_histogram(binwidth=0.25, position='dodge')
ggplot(Data, aes(x=RPG, fill=League)) + 
  geom_histogram(binwidth=0.25) + facet_grid(League ~ .)

## Scatterpoint
ggplot(Data, aes(x=AVG, y=RPG)) + geom_point()
ggplot(Data, aes(x=AVG, y=RPG, colour=League)) + theme_bw() + 
  scale_colour_brewer(palette="Set1") + geom_point(size=4)

## Simple box plots - you may need to 'reshape' the data for ggplot use
require(reshape)
M.Data <- melt(Data)
ggplot(M.Data, aes(x=variable, y=value)) + theme_bw() + geom_boxplot()
ggplot(subset(M.Data, variable %in% c("AVG", "OBP", "SLG")),
       aes(x=variable, y=value)) + theme_bw() + geom_boxplot()

## Box Plot with multiple variables
ggplot(Data, aes(x=factor(League), y=RPG, colour=League)) + theme_bw() + 
	scale_colour_brewer(palette="Set1") + geom_boxplot()
ggplot(Data, aes(x=factor(League), y=RPG, colour=League)) + theme_bw() + 
  scale_colour_brewer(palette="Set1") + geom_boxplot() + geom_point()
ggplot(Data, aes(x=factor(League), y=RPG, colour=League)) + theme_bw() + 
	scale_colour_brewer(palette="Set1") + geom_boxplot() + geom_jitter()

## Box plot + kernel density
ggplot(Data, aes(x=factor(League), y=RPG, colour=League)) + theme_bw() + 
	scale_colour_brewer(palette="Set1") + geom_violin()

## Plot points + model
ggplot(Data, aes(x=H, y=RPG)) + theme_bw() + geom_point() + 
	geom_abline(intercept = modelH$coef[1], slope= modelH$coef[2], colour="red") + 
	xlab("H") + ylab("Runs per Game")