Martin Frigaard 9/8/2017
First, load the tidyverse.
suppressWarnings(suppressMessages(library(tidyverse)))
suppressWarnings(suppressMessages(library(magrittr)))The tidyverse is a collection of R packages developed by RStudio’s Chief Scientist Hadley Wickham. These packages work well together as part of larger data analysis pipeline. To learn more about these tools and how they work together, read R for data science.
We will be using the bad_drivers data set from “Dear Mona, Which State Has The Worst Drivers?”. We load it into out working environment and give it a _df to identify that this object is the data frame.
## Observations: 51
## Variables: 8
## $ state <chr> "Alabama", "Alaska", "Arizona", "Arkansas", "Ca…
## $ num_drivers <dbl> 18.8, 18.1, 18.6, 22.4, 12.0, 13.6, 10.8, 16.2,…
## $ perc_speeding <int> 39, 41, 35, 18, 35, 37, 46, 38, 34, 21, 19, 54,…
## $ perc_alcohol <int> 30, 25, 28, 26, 28, 28, 36, 30, 27, 29, 25, 41,…
## $ perc_not_distracted <int> 96, 90, 84, 94, 91, 79, 87, 87, 100, 92, 95, 82…
## $ perc_no_previous <int> 80, 94, 96, 95, 89, 95, 82, 99, 100, 94, 93, 87…
## $ insurance_premiums <dbl> 784.55, 1053.48, 899.47, 827.34, 878.41, 835.50…
## $ losses <dbl> 145.08, 133.93, 110.35, 142.39, 165.63, 139.91,…
The aesthetic is the part of the graph we see. When we apply a numerical value (or variable) to an aesthetic, we refer to this as mapping. Consider the call below in which we map perc_speeding to insurance_premiums to the x and y. We assign this to a new object and add _aes to remind us this is the data frame and aesthetic mapping.
bad_drivers_df_aes <- bad_drivers_df %>% ggplot(aes(x = perc_speeding, y = insurance_premiums))
bad_drivers_df_aes
This creates a blank canvas (or carteisn coordinate system) with the two mapped variables (perc_speeding and insurance_premiums), but we don’t see any data points. These are added in the next layers.
We would like to see the scatter plot between these two variables, but in order to do that we need to understand something about the statistical transformations and geometric objects. Every geometric object has belongs to a particular statistical transformation, and every statistical transformation has a particular geometric object.
We can get under the hood of these functions by looking at the layer function.
p <- ggplot(data = diamonds, mapping = aes(x = carat))
p2 <- p + layer(
geom = "bar",
params = list(
binwidth = 0.5,
fill = "white",
color = "steelblue"
),
stat = "bin",
position = "identity"
)
p2
bad_drivers_df_aes + layer(
mapping = NULL, # already specified
data = NULL, # already specified
geom = "point",
stat = "identity",
position = "identity"
)
We can see "point" is the geom for a scatter plot, and that "identity" (which leaves the data unchanged) is the default stat and position. But we can lesson the typing by just using the shortcut geom_point().
The position argument adjusts the geometric objects to provide a clearer image. For example, I can add geom_jitter() to add some random noise to every data point. We will call this object gg_bad_drivers because it has all five layers.
gg_bad_drivers <- bad_drivers_df_aes_stat_geom +
geom_jitter(width = 10, height = 10)
gg_bad_drivers
This makes more sense when we are dealing with categorical or discrete values and many data points overlap.
Now that we have the basics down, we can start adding more layers to build better graphs.
Lets look at the relationship between perc_alcohol (the percentage of drivers involved in fatal collisions who were alcohol-impaired) and losses (financial losses incurred by insurance companies for collisions per insured driver), but add the color aesthetic.
The legend makes the plot hard to see, so I will use the theme(legend.position = "none") function to remove the legend altogether.
bad_drivers_df %>% ggplot(aes(x = perc_alcohol, y = losses)) +
geom_point(aes(color = state)) +
theme(legend.position = "none")
That’s better. But I want to the point to represent the number of num_drives (the number of drivers involved in fatal collisions per billion miles).
bad_drivers_df %>% ggplot(aes(x = perc_alcohol, y = losses)) +
geom_point(aes(color = state, size = num_drivers)) +
theme(legend.position = "none")
It’s a little hard to tell which state is which without labels, so I’ll add a geom_text and apply the size of the text to num_drivers.
bad_drivers_df %>% ggplot(aes(x = perc_alcohol, y = losses)) +
geom_point(aes(color = state)) +
geom_text(aes(label = state, color = state, size = num_drivers)) +
theme(legend.position = "none")
Lets break this down further by region. Below are vectors for each US state region.
american_midwest <- c("Illinois", "Indiana", "Iowa", "Kansas", "Michigan",
"Minnesota", "Missouri", "Nebraska", "North Dakota",
"Ohio", "South Dakota", "Wisconsin")
american_northeast <- c("Connecticut", "Maine", "Massachusetts",
"New Hampshire", "Rhode Island", "Vermont",
"New Jersey", "New York", "Pennsylvania")
american_south <- c("Delaware", "Florida", "Georgia", "Maryland",
"North Carolina", "South Carolina", "Virginia",
"District of Columbia", "West Virginia", "Alabama",
"Kentucky", "Mississippi", "Tennessee", "Arkansas",
"Louisiana", "Oklahoma", "Texas")
american_west <- c("Arizona", "Colorado", "Idaho", "Montana", "Nevada",
"New Mexico", "Utah", "Wyoming", "Alaska", "California",
"Hawaii", "Oregon", "Washington")I want to create a new variable region, in the bad_drivers_df data frame, and then facet the graphs according to this new variable.
Let’s do this is one pipeline
bad_drivers_df %>%
mutate(
region =
case_when(
state %in% american_midwest ~ "midwest",
state %in% american_northeast ~ "northeast",
state %in% american_south ~ "south",
state %in% american_west ~ "west",
TRUE ~ "other"
)
) %>%
ggplot(aes(x = perc_alcohol, y = losses)) +
geom_point(aes(color = region, size = num_drivers)) +
geom_text(aes(label = state, size = 8)) +
facet_grid(. ~ region)
This graph combines the positions, colors, size, labels, and facets to display the relationship between the amount of money lost by insurance companies for collisions per insured driver and the percentage of alcohol-impaired drivers involved in fatal collisions.
Now that we know the basics for a ggplot2 graph, we can start to build on these foundations to get more interesting representations of the data.
We will start with the native diamonds data set from the ggplot2 package.
I use gg for data sets I create visualizations with (for pattern matching purposes).
## Observations: 53,940
## Variables: 10
## $ carat <dbl> 0.23, 0.21, 0.23, 0.29, 0.31, 0.24, 0.24, 0.26, 0.22, 0.23,…
## $ cut <ord> Ideal, Premium, Good, Premium, Good, Very Good, Very Good, …
## $ color <ord> E, E, E, I, J, J, I, H, E, H, J, J, F, J, E, E, I, J, J, J,…
## $ clarity <ord> SI2, SI1, VS1, VS2, SI2, VVS2, VVS1, SI1, VS2, VS1, SI1, VS…
## $ depth <dbl> 61.5, 59.8, 56.9, 62.4, 63.3, 62.8, 62.3, 61.9, 65.1, 59.4,…
## $ table <dbl> 55, 61, 65, 58, 58, 57, 57, 55, 61, 61, 55, 56, 61, 54, 62,…
## $ price <int> 326, 326, 327, 334, 335, 336, 336, 337, 337, 338, 339, 340,…
## $ x <dbl> 3.95, 3.89, 4.05, 4.20, 4.34, 3.94, 3.95, 4.07, 3.87, 4.00,…
## $ y <dbl> 3.98, 3.84, 4.07, 4.23, 4.35, 3.96, 3.98, 4.11, 3.78, 4.05,…
## $ z <dbl> 2.43, 2.31, 2.31, 2.63, 2.75, 2.48, 2.47, 2.53, 2.49, 2.39,…
We will start by breaking down the components of a basic ggplot call. Before we do this, I want to get some basic summary statistics for price and carat.
knitr::kable(
dmnds_gg %>%
dplyr::select(carat) %>%
summarise(
min_carat = min(carat),
max_carat = max(carat),
med_carat = median(carat),
mean_carat = mean(carat)
)
)| min_carat | max_carat | med_carat | mean_carat |
|---|---|---|---|
| 0.2 | 5.01 | 0.7 | 0.7979397 |
NOTE: Makes tables pretty using knitr::kable()
knitr::kable(
dmnds_gg %>%
dplyr::select(price) %>%
summarise(
min_price = min(price),
max_price = max(price),
med_price = median(price),
mean_price = mean(price)
)
)| min_price | max_price | med_price | mean_price |
|---|---|---|---|
| 326 | 18823 | 2401 | 3932.8 |
Ok now that we have an idea for each of these variables individually, lets see how they looks in a graph plotted against each other.
What do we see? This graph looks like a scatter plot between price and caret. There seems to be a pattern in this relationship (which we will discuss more later), but is this what we would expect to see?
We tend to think of graphics and visualizations in terms of the shapes we see (i.e. the dots, bars, lines, etc.). The ggplot() function creates plots in R conveniently named after these geometric objects using a geom_ (i.e. geom_bar for bar charts, geom_line for line charts, etc). Each geom_ argument has five components:
data: the data set (diamonds)mapping: horizontal (x) position = carat, vertical (y) position = pricegeom: geometric objects like scatter plots, histograms, lines, etc. = pointstat: statistical transformations (i.e. counts or bins) = identityfor raw dataposition: adjusting any overlapping objects = identityfor raw dataThe grammar allows us to build a plot with each component explicitly using the layer function (and the native diamonds data set).
I like to capitalize on English’s flat branching structure, because each clause (graph component) ends up on it’s own line, and the entire graph command can extended diagonally with commas(,). I don’t always remember to arrange it this way, but it helps when I am building a plot or trying to read someone else’s.
ggplot() +
layer(
data = diamonds, # 1 THE DATA SET
mapping = aes( # 2 MAPPINGS
x = carat, # x axis
y = price
), # y axis
geom = "point", # 3 GEOM_FUNCTION
stat = "identity", # 4 STAT
position = "identity"
) # 5 POSITION
The reason we don’t have to state each part explicitly is the beauty of the ggplot2 package. The grammar comes loaded with smart defaults for many of the graph components, so we don’t have to remember what stat goes with each geom.
There are actually seven parts to each ggplot2 graph, although not all are required. We will explore the other functions later. A template for the grammar is available below:
ggplot(data = <THE DATA SET>) + # default data and aesthetic mapping
geom_<GEOM_FUNCTION>( # the geometric object
mapping = aes(<MAPPINGS>), # map aesthetics (with scales)
stat = <STAT>, # statistical transformations
position = <POSITION> # position adjustments
) +
<COORDINATE_FUNCTION> + # coordinate systems
<FACET_FUNCTION> # facets It’s also available here
For additional tidyr functions using qplot, we will use the data for the story “A Statistical Analysis of the Work of Bob Ross”, available here. Let’s load it in and take a look.
## Observations: 403
## Variables: 71
## $ episode <chr> "S01E01", "S01E02", "S01E03", "S01E04", "S01E05"…
## $ season <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, …
## $ episode_num <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 1, 2,…
## $ title <chr> "A WALK IN THE WOODS", "MT. MCKINLEY", "EBONY SU…
## $ apple_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ aurora_borealis <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ barn <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ beach <int> 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, …
## $ boat <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ bridge <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ building <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ bushes <int> 1, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, …
## $ cabin <int> 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ cactus <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ circle_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ cirrus <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, …
## $ cliff <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ clouds <int> 0, 1, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, …
## $ conifer <int> 0, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 0, …
## $ cumulus <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, …
## $ deciduous <int> 1, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 0, 1, …
## $ diane_andre <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ dock <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ double_oval_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ farm <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ fence <int> 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, …
## $ fire <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ florida_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ flowers <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ fog <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ framed <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ grass <int> 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, …
## $ guest <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ half_circle_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ half_oval_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ hills <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ lake <int> 0, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, …
## $ lakes <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ lighthouse <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ mill <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ moon <int> 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ mountain <int> 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 0, …
## $ mountains <int> 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, …
## $ night <int> 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ ocean <int> 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, …
## $ oval_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ palm_trees <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ path <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ person <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ portrait <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ rectangle_3d_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ rectangular_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ river <int> 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ rocks <int> 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ seashell_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ snow <int> 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, …
## $ snowy_mountain <int> 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, …
## $ split_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ steve_ross <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ structure <int> 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ sun <int> 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, …
## $ tomb_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ tree <int> 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, …
## $ trees <int> 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, …
## $ triple_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ waterfall <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ waves <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, …
## $ windmill <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ window_frame <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ winter <int> 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
## $ wood_framed <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, …
Ugh–what IS this? Well, this is a standard binary format data set, with each variable telling us the presence (1) or absence (0) of each variable.
NOTE: if you are ever creating binary variables, always have 1 equal the presence of the variable name, and 0 coded as the absence. For example, a binary gender variable could be named male, and 1 could be the numerical code for male gender (this also helps with sorting because the character variable will sort female before male, which aligns with the 0 and 1). Developing these habits early (and sticking to them!) makes your analyses more organized and easy to follow.
We will convert this format into a tidy data set.
bob_tidy_gg <- bob_gg %>%
gather(object, present, -c(episode, season, episode_num, title)) %>%
mutate(present = as.logical(present)) %>%
arrange(episode, object)
bob_tidy_gg %>% glimpse()## Observations: 27,001
## Variables: 6
## $ episode <chr> "S01E01", "S01E01", "S01E01", "S01E01", "S01E01", "S01E…
## $ season <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ episode_num <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…
## $ title <chr> "A WALK IN THE WOODS", "A WALK IN THE WOODS", "A WALK I…
## $ object <chr> "apple_frame", "aurora_borealis", "barn", "beach", "boa…
## $ present <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, TRUE, …
The ggplot2 package (and all others in the tidyverse) work best with tidy data. If you are ever curious if your data is tidy, consider it’s structure and measurements. The original bob_ross data set had 403 observations and 71 variables. This new bob_tidy_gg has 27,001 observations and 6 variables. That’s because the original bob_ross data set had variables for every item his paintings, and whether it was there (or not). This doesn’t meet the tidy rule of one variable per column, because these columns are actually representing a measurement of painting objects, and whether the object is present or not. In this case, we prefer the data structure that contains the most granular measurement variables.
## Parsed with column specification:
## cols(
## id = col_double(),
## firstname = col_character(),
## midname = col_character(),
## lastname = col_character(),
## fullname = col_character(),
## ftin = col_character(),
## feet = col_double(),
## inches = col_double(),
## meters = col_double(),
## gender = col_character()
## )
## Observations: 5,502
## Variables: 10
## $ id <dbl> 1, 8, 9, 2, 3, 4, 5, 6, 7, 10, 17, 18, 19, 20, 32, 33, 34…
## $ firstname <chr> "Verne", "Herve", "David", "Tony", "Warwick", "Phil", "Mi…
## $ midname <chr> NA, NA, NA, NA, NA, NA, "J", NA, NA, NA, NA, NA, NA, NA, …
## $ lastname <chr> "Troyer", "Villechaize", "Rappaport", "Cox", "Davis", "Fo…
## $ fullname <chr> "Verne Troyer", "Herve Villechaize", "David Rappaport", "…
## $ ftin <chr> "2ft 8in", "3ft 10in", "3ft 11in", "3ft 6in", "3ft 6in", …
## $ feet <dbl> 2, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, …
## $ inches <dbl> 8.0, 10.0, 11.0, 6.0, 6.0, 6.0, 7.0, 8.0, 9.0, 1.7, 10.0,…
## $ meters <dbl> 0.81280, 1.16840, 1.19380, 1.06680, 1.06680, 1.06680, 1.0…
## $ gender <chr> "M", "M", "M", "M", "M", "M", "M", "M", "M", "F", "F", "F…
For example, let’s say I want to look at the relationship in child and parent heights in the HistData::PearsonLee dataset. This data set is a little different than most because it has a frequency variable.
library(HistData)
hts_df <- PearsonLee %>% tbl_df()
hts_df %>%
ggplot(aes(x = parent, y = child)) +
geom_point()
| | min | Q1 | median | Q3 | max | mean | sd | n | missing | | | --: | ---: | -----: | ---: | ----: | ------: | -------: | ---: | ------: | | | 2.8 | 56.1 | 72.7 | 88.9 | 230.7 | 70.9818 | 29.12536 | 9922 | 78 |
Get the summary for Weight.
| | min | Q1 | median | Q3 | max | mean | sd | n | missing | | | ---: | ----: | -----: | ----: | ----: | -------: | -------: | ---: | ------: | | | 83.6 | 156.8 | 166 | 174.5 | 200.4 | 161.8778 | 20.18657 | 9647 | 353 |
Now with a little bit of summary info I can start to imagine what this scatter plot will look like. First, my x and y scales will go from 2.8 - 230.7 and 83.6 - 200.4. I also know that most of the data (dots) will be around 70 for weight and 160 for height. Does this tell me if the relationship between them is linear?
If this is a linear relationship, I should expect to see a bunch of dots grouped together on the diagonal (lower-left to upper-right). I am going to create a scatter plot with these two variables, but first I am going to filter out the missing Height and Weight values.
nhanes %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point()
The relationship between these variables looks fairly linear (i.e. diagonal), but it also seems to have a bit of a curve. We also have a fair amount of overplotting (points sitting on top of one another). We can clean this up in a couple different ways. The first is to impose a size value in the geom_point().
nhanes %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point(size = 0.6)
The size argument is in reference to the size of the points. We could also change the tranparency of the points using the alpha aesthetic.
nhanes %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point(alpha = 1 / 10)
Finally, we could also not use all of the data (i.e. take a random sample of 50 observations from the nhanes data set)
nhanes %>%
sample_n(size = 500) %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point()
All of these solutions work, but I would use the size or alpha options because we don’t have to choose how many observations to keep.
So does this relationship look linear? We can use the geom_smooth() function to see what line best fits these data. We’ll reduce the tranparency to 1/10 for clarity, too.
nhanes %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point(alpha = 1 / 10) +
geom_smooth()## `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'
It looks like the geom_smooth() function fit a line using method = "gam", and not method = "lm". This is another default in ggplot2–plots with 1000+ observations use the "gam" method. We can specify the linear method by simply adding another geom_smooth(), but we should make sure to specify a different line color (color = "red").
nhanes %>%
filter(!is.na(Weight) & !is.na(Height)) %>%
ggplot(aes(x = Weight, y = Height)) +
geom_point(alpha = 1 / 10) +
geom_smooth() +
geom_smooth(method = "lm", color = "red")## `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'
NOTE: the alpha aesthetic can take fractions (1/10) or decimals (0.1).
Scatter plots (geom_point()) are a great way to show the relationship data from two continuous variables. But what if you want to look at how a continuous variable is distributed across the levels of a categorical (or factor) variable? There is more than one way to do this, but we will start with a stacked bar chart.
I want to know how the variable Age varies across Gender in the nhanes data set. I can get some quick summary statistics by using group_by from dplyr:
nhanes %>%
group_by(Gender) %>%
summarise(
min = min(Age),
max = max(Age),
median = median(Age),
mean = mean(Age),
Mode = mode(Age)
)## # A tibble: 2 x 6
## Gender min max median mean Mode
## <fct> <int> <int> <dbl> <dbl> <chr>
## 1 female 0 80 37 37.6 numeric
## 2 male 0 80 36 35.8 numeric
Ok my summary statistics let me know that I should be expecting a graph that ranges from 0 to 80 on the x. Can we tell what the most common age is? The statistical term for this is the mode, and unfortunately R doesn’t have a quick and easy way to get the mode for variables in data frames. No problem though
my_mode <- function(x, na.rm = FALSE) {
if (na.rm) {
x <- x[!is.na(x)]
}
ux <- unique(x)
return(ux[which.max(tabulate(match(x, ux)))])
}nhanes %>%
group_by(Gender) %>%
summarise(
min = min(Age),
max = max(Age),
median = median(Age),
mean = mean(Age),
mode = my_mode(Age)
)## # A tibble: 2 x 6
## Gender min max median mean mode
## <fct> <int> <int> <dbl> <dbl> <int>
## 1 female 0 80 37 37.6 80
## 2 male 0 80 36 35.8 80
A stacked bar chart uses the fill aesthetic to separate the levels of a categorical or factor variable by color.
The y axis is a count of number of observations in each bar. ggplot2 also automatically provides a legend to keep track of each level.
This was a very brief tutorial of the ggplot2 package, so I recommend learning more about the package by typing library(help = "ggplot2") into your R console, checking out the ggplot2 tidyverse page, or purchasing the ggplot2 book.