---
title: "EDA"
author: "JJB + Course"
date: "03/04/2019"
output:
   html_document:
     toc: true
     toc_float:
       collapsed: false
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, cache = TRUE)
library("ggplot2")
theme_set(theme_bw())
```

# Exploratory Data Analysis (EDA)

## Example: 5 Stat Summary vs. Graph

```{r eda}

summary(iris[,4:5])

ggplot(iris) +
  geom_point(
    aes(x = Petal.Width,
        y = Petal.Length,
        color = Species))
```


## Example: Patterns

```{r gen-lin, echo = FALSE}
# Create a sequence
start = 1
end = 100
x = seq(start, end, by = 2)

# Store into data.frame
df = data.frame(x = x, 
                y = x)

# Plot result
ggplot(df) +
  geom_point(aes(x,y)) +
  labs(title = "Linear Relationship")
```


```{r gen-wobbly-lin, echo = FALSE}
# Create a sequence
start = 1
end = 100
x = seq(start, end, by = 2)

# Add some noise to original signal
df = data.frame(x = x, 
                y = 2*x + 1 + rnorm(length(x), mean = 3, sd = 15)
               )

# Plot result
ggplot(df) +
  geom_point(aes(x,y)) +
  labs(title = "Fuzzed Linear Relationship")
```


## Example: Grains of Sand

As an example of why graphs are helpful, consider this generated data set and accompanying quantitative EDA.

```{r grains-of-sand, cache = TRUE}
set.seed(2018)                                    # Set seed for reproducibility
n = 1e4                                           # Number of observations

a = matrix(rnorm(n * 2), ncol = 2)                # Generate values 
rng = runif(n, 0, 2 * pi) 
b = 0.5 * cbind(sin(rng), cos(rng))

o = rbind(a, b)                                   # Combine generate data

my_data = as.data.frame(o[sample(nrow(o)), ])     # Randomly sample data

colnames(my_data) = c("x", "y")
```

## Example: Quantitative EDA

```{r grains-summary, cache = TRUE, dependson="grains-of-sand"}
dim(my_data)                # Dimensions of Data
nrow(my_data)               # Number of Rows
ncol(my_data)               # Number of Columns

summary(my_data)            # Summarize data

head(my_data)               # First observations in data
tail(my_data)               # Last observations in data

class(my_data)              # Kind of data
str(my_data)                # Layout of data

head(is.na(my_data))        # Any missing values?
```


## Example: Visual EDA

```{r grains-bad-plot, cache = TRUE, dependson="grains-of-sand"}
library("ggplot2")
theme_set(theme_bw())
ggplot(my_data) + 
   geom_point(aes(x = x, y = y))
```

```{r grains-redux, cache = TRUE, dependson="grains-of-sand"}
ggplot(my_data) + 
   geom_point(aes(x = x, y = y), alpha = 0.05)
```

# Example: Categorical vs. Continuous

```{r cat-vs-cont}
ggplot(iris) + 
  aes(x = Species, fill = Species) +
  geom_bar() + 
  labs(title = "1 Variable: Categorical Counts")

ggplot(iris) + 
  aes(x = Petal.Width, fill = Species) + 
  geom_histogram() + 
  labs(title = "1 Variable: Continuous Binning")
```

## Example: Covariation

```{r cov-ggplot2}
ggplot(iris) +
  aes(Species, Petal.Width, color = Species) + 
  geom_count() +
    ggtitle("Covariation: Petal.Width vs. Species")
```

# Graph Systems in R

## Base _R_ 

```{r base-r-comparison}
plot(iris$Sepal.Length, iris$Petal.Length,        # x variable, y variable
     col = iris$Species,                          # Color points by species
     pch = 15,                                    # Kind of point to use
     cex = 1,                                     # Size of points
     xlab = "Sepal Length",                       # Label for x-axis
     ylab = "Petal Length",                       # Label for y-axis 
     main = "Visualizing Sepal and Petal Length") # Main title of plot

legend (x = 4.2, y = 7,                           # Manually specify legend
        legend = levels(iris$Species), 
        col = c(1, 2, 3), pch = 15)
```


```{r lattice-comparison}
library("lattice")
xyplot(
  Petal.Length ~ Sepal.Length,                 # y variable, x variable
  groups = Species,                            # Color by species
  main = "Visualizing Sepal and Petal Length", # Graph title
  data = iris,                                 # Data set
  auto.key = TRUE                              # Make legend
  ) 
```

```{r ggplot2-comparison}
library("ggplot2")
ggplot(data = iris) + 
  aes(x = Sepal.Length, y = Petal.Length, color = Species) + 
  geom_point() + 
  labs(title = "Visualizing Sepal and Petal Length") + theme_gray() + 
  theme()
```


# ggplot2

## Example: Adding of a Layer

```{r start-ggplot}
g = ggplot()
g
```

```{r preview-data, warning = FALSE}
# Take a look at the data
head(mpg)

class(mpg$displ)
class(mpg$hwy)
```

Add data and a mapping

```{r add-mappings}
g = ggplot(data = mpg) +
     aes(x = displ, y = hwy, colour = class)
g
```

Display data using a geometric point.

```{r add-points}
g = g + geom_point()
g
```

Add trends of data per subgroup using a smoothing function.

```{r add-trend-lines}
g = g + geom_smooth()
g
```

## Example: Layer Details

```{r view-layers}
names(g)
str(g)
```

## Example: Graph Labels

```{r graph-labels}
g = ggplot(mpg) +
    aes(x = displ, y = hwy, color = class) +
    geom_point() + 
    labs(title = "Engine Displacement (Liters) vs. Highway MPG", 
             subtitle = "Based on data from 1999 to 2008",  
             x = "Engine Displacement (Liters)", 
             y ="Highway Miles Per Gallon", 
             color = "Type of Car")
g
```


### Example: Add labels to the points:

```{r add-labels-plots}
g +
  geom_text(aes(label = rownames(mpg)))
```


## Example: Exporting a `ggplot2` Object

```{r cli-export, eval = FALSE}
# Save plot into object
g = ggplot(iris) + 
    aes(x = Sepal.Length, y = Petal.Length, colour = Species) + 
    geom_point() + theme_bw()

g

# Call ggsave to render plot to file
ggsave("Rplot.png", plot = g)

ggsave("Rplot.png", plot = g, width = 10)
```


# Making Graphs

We will be using the `ucidata` package to create plots

```{r}
# install.packages("ucidata")
# devtools::install_github("coatless/ucidata")
library("ucidata")

head(bike_sharing_daily)

# ?bike_sharing_daily   # View help documentation for variable information
```


## Example: Scatterplot

```{r}
theme_set(theme_bw())

ggplot(data = bike_sharing_daily) +
  geom_point( mapping = aes( x = dteday, y = cnt) )
```

## Example: Coloring Groups

```{r}
ggplot(data = bike_sharing_daily) +
  geom_point( mapping = 
                aes( x = dteday, y = cnt,
                     color = season) )
```

## Example: Transparency

```{r}
ggplot(data = bike_sharing_daily, 
       mapping = aes( x = dteday, y = cnt, color = season)) +
    geom_point( alpha = 1)

ggplot(data = bike_sharing_daily, 
       mapping = aes( x = dteday, y = cnt, color = season)) +
    geom_point( alpha = 0.5)


ggplot(data = bike_sharing_daily, 
       mapping = aes( x = dteday, y = cnt, color = season)) +
    geom_point( alpha = 0.25)

```


## Example: Faceting a plot

```{r}
ggplot(data = bike_sharing_daily, 
       mapping = aes( x = dteday, y = cnt, color = season)) +
    geom_point( ) + 
    facet_wrap(~weekday)
```

```{r}
model = lm(cnt ~ season + weekday, data = bike_sharing_daily)

X_design = model.matrix(cnt ~ season + weekday, data = bike_sharing_daily)

summary(model)
```


### Exercise: Changing `x` and `y`


```{r}
# Specifying aesthetic on only one geometric object.
ggplot(data = bike_sharing_daily) +
  geom_point(mapping = aes(x = dteday, y = atemp, color = season))
```

```{r}
# You can avoid having to respecify the aesthetic for the graph
ggplot(data = bike_sharing_daily) +
  aes(x = dteday, y = atemp, color = season) + 
  geom_point() 
```

```{r}

```

```{r}

```


## Example: Line Graph

```{r}
ggplot(data = bike_sharing_daily) +
  aes( x = dteday, y = cnt) + 
  geom_line( )
```

### Exercise: Changing plots

```{r}

```


## Example: Barplot

```{r demo-barplot}
ggplot(data = bike_sharing_daily) +
  aes( x = workingday) +
  geom_bar()
```

Tabulate categorical frequencies for **one** variable

```{r calculate-frequencies-one-var}
table(bike_sharing_daily$workingday)

table(bike_sharing_daily$weathersit)
```

Tabulate categorical frequencies for **two** variable

```{r calculate-frequencies-two-var}
table(bike_sharing_daily$weathersit, bike_sharing_daily$holiday)

table(bike_sharing_daily$workingday, bike_sharing_daily$holiday)
```

```{r show-freq-on-barplot}
ggplot(data = bike_sharing_daily) +
  aes( x = holiday) + 
  geom_bar() + facet_wrap(~weathersit)
```

## Example: Stacked Barplot

```{r stacked-bar-plot}
ggplot(data = bike_sharing_daily) +
  aes( x = mnth , y = cnt) + 
  geom_bar( stat = "identity" )
```

```{r stacked-bar-plot-color}
ggplot(data = bike_sharing_daily) +
  aes( x = mnth , y = cnt, fill = weekday) + 
  geom_bar( stat = "identity" )
```

```{r demo-stacked-bar-plot-facet}
ggplot(data = bike_sharing_daily) +
  aes(x = mnth, y = cnt, fill = weekday) + 
  geom_bar( stat = "identity") +
  facet_wrap(~yr) + 
  theme(axis.text.x = element_text(angle = 75, hjust = 1))
```


## Example: Histogram

```{r demo-histogram}
ggplot(data = bike_sharing_daily) +
  aes( x = cnt ) + 
  geom_histogram(  )
```

Change binwidth! 

```{r histogram-demo-binwidth-high}
ggplot(data = bike_sharing_daily) +
  aes( x = cnt ) + 
  geom_histogram( binwidth = 100 )

ggplot(data = bike_sharing_daily) +
  aes( x = cnt ) + 
  geom_histogram( binwidth = 250 )
```


```{r histogram-demo-binwidth-low}
ggplot(data = bike_sharing_daily) +
  aes(x = atemp, fill = season) +
  geom_histogram(binwidth = 0.05)

ggplot(data = bike_sharing_daily) +
  aes(x = atemp, fill = season) +
  geom_histogram(binwidth = 0.05)
```

## Example: Boxplot

```{r boxplot-demo}
ggplot(data = bike_sharing_daily) +
  aes( x = weekday, y = cnt, color = weekday) +
  geom_boxplot()
```

## Example: Boxplot with Points

```{r boxplot-with-points}
ggplot(data = bike_sharing_daily) + 
  aes( x = weekday, y = cnt, color = weekday) +
  #geom_boxplot() +
  geom_point()
```

## Example: Boxplot with Jittered Points


```{r boxplot-with-jitter}
ggplot(data = bike_sharing_daily) +
  aes( x = weekday, y = cnt, color = weekday) +
  geom_boxplot() + geom_jitter()
```

## Example: Facetted Boxplot

```{r facetted-boxplot}
ggplot(data = bike_sharing_daily) +
  aes( x = weekday, y = cnt, color = weekday) + 
  geom_boxplot() + facet_wrap(~season)
```


## Example: Violin Plot

```{r violin-plot-with-jitter}
ggplot(data = bike_sharing_daily) + aes( x = weekday, y = cnt, color = weekday) +
  geom_violin() + geom_jitter(height = 0)
```

# Themes

## Example: Modifying a Theme

```{r modify-theme-temporary}
g = ggplot(data = iris) + 
  geom_point(aes(Sepal.Length, Petal.Length, color = Species)) + 
  labs(title = "Visualizing Sepal and Petal Length") 

g 

g + theme(
  panel.background = element_blank(),
  axis.text = element_blank()
)
```


```{r themes-predefined, fig.width = 6, fig.height = 2.5}
g = ggplot(mtcars) +
      geom_point(aes(x = wt, y = mpg,
                     color = factor(gear))) + 
      facet_wrap(~am)

# Default theme:
g + theme_gray()

# Black and white
g + theme_bw()

# Dark
g + theme_dark()

# Classic
g + theme_classic()

# Geoms only
g + theme_void()
```