R Introduction.Rpres

R Introduction
========================================================
author: Sixiang Hu
width: 1280
height: 1024
autosize: true

What is R?
========================================================
- Open source : lots of experts in different areas are contributing to this software

- Package based : lots of packages available and you can create your own.

- Free

```{r setup,echo=FALSE,message=FALSE}
#https://github.com/mages/googleVis_KoelnR_May_2014/blob/master/googleVis_update.Rpres
options(width=120)
opts_chunk$set(out.width="1000px", fig.align="center")
library(googleVis)
options(gvis.plot.tag="chart")
library(ggplot2)
library(ggthemes)
library(forecast)
library(xts)
library(fractal) 
```

Getting Started
========================================================
* Download R 3.1.2 (Pumpkin Helmet):
  + http://cran.r-project.org/

* Download RStudio (Optional):
  + http://www.rstudio.com/products/rstudio/download/
  + To get a beautiful user interface
  + Easier maintains

Basic Assignment and Operations
========================================================
* Arithmetic Operations:
```{r}
1+2*exp(3)/sin(4)
```

* Assignment:
```{r}
a <- runif(4)
a
```

* Use help in R:
```{r}
?sin
```

Variables in R
========================================================
There are several types of variables can be used in R:
integer, numerical, character, date, vector, matrix, list, 
and data frame.

```{r}
a <- sample(LETTERS,5,replace=T)
a
class(a)
class(data.frame(c(a=a,b=a)))
```

Data Frame
========================================================
Data frame is the widest used data type used in R.
It can have different type of value in different columns, but 
values in a column must keep the same.

```{r}
a <- c(1,2,3)
b <- c("a","b","c")
df <- data.frame(a=a,b=b)
df
df$a
df[2,2]
```

List
========================================================
List can be used to store different types of variables with 
different length.

```{r}
a <- c(1,2,3)
b <- c("a","b","c")
lt <- list(a,b)
lt
lt[[2]]
```

Function
========================================================
We can create functions that can be used repetitively:
```{r}
my_square <- function(x) {x*x}
my_square(5)
```

Plot
========================================================
```{r PlotCars,fig.width=12, fig.height=12, fig.show='hold', fig.align='center'}
par(mfrow=c(2,2))
plot(cars,type="l")
hist(cars$speed)
plot(cars)
hist(cars$dist)
```

Modelling
========================================================
```{r}
 model1 <- glm(mpg~gear+cyl,data=mtcars,family=Gamma)
summary(model1)
```

We can use different family and link function for GLM 
modelling.  Interactions, offset, splines and transformation 
are all can be done within R.

For some large dataset (>100MB), it will be slow for a fit.
Some packages `bigGLM` can be used to accelerate to process.

Advanced Plot (ggplot2)
========================================================
```{r}
ggplot(data = msleep, aes(x = log(bodywt), y = sleep_total)) +
  geom_point(aes(shape=vore), size=3)+
  theme_wsj()
```

GoogleVis Chart
========================================================
```{r GoogleVisChart, results='asis', tidy=FALSE}
df <- data.frame(country=c("US", "GB", "BR"),val1=c(1,3,4),val2=c(23,12,32) )
Bar1 <- gvisBarChart(df, xvar="country", yvar=c("val1", "val2"))
plot(Bar1)
```

Example 1: Time Series for Market Trending
========================================================
forecast package gives automaticallal fitting to a time series developed by [Rob Hyndman](http://robjhyndman.com/) 
```{r TSCode1}
MSM.raw <- read.csv("U:\\Steven\\MSMIQ\\2014 MarketPremall CREATION_DATE AGE_BAND.csv",
                    head=T,sep=",",stringsAsFactors=FALSE)
MSM.raw$TIME <- as.Date(MSM.raw$CREATION_DATE,"%d%b%Y")
MSM.raw <- subset(MSM.raw,CREATION_DATE != "" &
                    AGE_BAND == "" & 
                    TIME>"2013-07-01" & TIME < "2014-11-30")
MSM.t<- as.xts(MSM.raw$MARKETPRICETOP5ANNUAL,MSM.raw$TIME,start=c(2013,180),end=c(2014,333),f=1/365)
obs <-length(MSM.t)
#Automatically fitting an ARIMA time series model:
MSM.t.Arima <- auto.arima(MSM.t,stationary=FALSE, seasonal=TRUE, xreg=1:obs,allowdrift=TRUE,parallel=TRUE,stepwise=FALSE)
summary(MSM.t.Arima)
```

```{r, echo=FALSE}
#We can test the stationary of a time series object:
#Staionarity Test, T: Variation over time, H0: Stationarity
#MSM.test <- stationarity(MSM.t)
#MSM.test
```

Example 1: Time Series for Market Trending (Continuous)
========================================================
```{r PlotTS,fig.width=12, fig.height=12, fig.show='hold', fig.align='center'}
MSM.t.forecast<-forecast(MSM.t.Arima,xreg=(obs+1):(obs+7))
plot(MSM.t.forecast)
```

Example 2: Chainladder for Reserving
========================================================
```{r}
```

Example 3: Decision Tree (Data Mining)
========================================================
```{r}
```
Example 4: Geographical Analysis
========================================================
```{r}
```
Further Online Tutorial
========================================================
* R Tutorial (Statistics)
  + http://www.r-tutor.com/

* CRAN R Documentations
  + http://cran.r-project.org/

* Data Camp R
  + https://www.datacamp.com/
  
* Kaggle
  + https://www.kaggle.com/