Week 7: Code along 6

R for Data Science: Chapter 6

Introduction

Questions

Variation

Visualizing distributions

diamonds %>%
  ggplot(aes(x = cut)) +
  geom_bar()

diamonds %>%
  ggplot(mapping = aes(x = carat)) + 
  geom_histogram(binwidth = 0.5)

diamonds %>%
  
  filter(carat < 3) %>%
  
  ggplot(aes(x = carat)) +
  geom_histogram(binwidth = 0.5)

diamonds %>%
  ggplot(aes(x = carat, color = cut)) +
  geom_freqpoly()

## `stat_bin()` using `bins = 30`. Pick better value `binwidth`.

Typical values

diamonds %>%
  
  # Filter out diamonds > 3 carat
  filter(carat < 3) %>%
  
  # Plot
  ggplot(aes(x = carat)) +
  geom_histogram(binwidth = 0.01)

faithful %>%
  ggplot(aes(eruptions)) +
  geom_histogram(binwidth = 0.25)

Unusual values

diamonds %>%
  ggplot(aes(y)) +
  geom_histogram()

## `stat_bin()` using `bins = 30`. Pick better value `binwidth`.

diamonds %>%
  ggplot(aes(y)) +
  geom_histogram() +
  coord_cartesian(ylim = c(0, 50))

## `stat_bin()` using `bins = 30`. Pick better value `binwidth`.

Missing Values

diamonds %>%
  
  # filter(y < 3 | y> 20) %>%
  
  mutate(y = ifelse(y < 3 | y > 20, NA, y)) %>%
  
  # Plot
  ggplot(aes(x = x, y = y)) +
  geom_point()

## Warning: Removed 9 rows containing missing values or values outside the scale range
## (`geom_point()`).

Covariation

A categorical and continous variable

diamonds %>%
  
  ggplot(aes(x = cut, y = price)) +
  geom_boxplot()

Two categrical variables

diamonds %>% 
  
  count(color, cut) %>%  
  
  ggplot(aes(x = color, y = cut, fill = n)) +
  geom_tile()

Two continous variables

library(hexbin)
diamonds %>%
  ggplot(aes(x = carat, y = price)) +
  geom_hex()

diamonds %>%
  filter(carat < 3) %>%
  ggplot(aes(x = carat, y = price)) +
  geom_boxplot(aes(group = cut_width(carat, 0.1)))

## Warning: Orientation is not uniquely specified when both the x and y aesthetics are
## continuous. Picking default orientation 'x'.

patterns and models

library(modelr)
mod <- lm(log(price) ~ log(carat), data = diamonds)

diamonds4 <- diamonds %>%
  modelr::add_residuals(mod) %>%
  mutate(resid = exp(resid))

diamonds4 %>%
  ggplot(aes(carat,resid)) +
  geom_point()

diamonds4 %>%
  ggplot(aes(cut, resid)) +
  geom_boxplot()