Our Voting Results
- Disclaimer: I teach Stat 100 right after Stat 108 and I think a Stat 100 student took off with half of our graphs!
- Update: Found them...
Recall: The Grammar of Graphics
data: dataset that contains the data
geom: geometric shape that the data are mapped to
- point, line, bar, text, ...
aesthetic: visual properties of the geom
- x position, y position, color, fill, shape
coord: coordinate system
- Cartesian, polar, geographic
scale: controls how data are mapped to the visual values of the aesthetic
- EX: particular colors, linear
guide: legend to help user convert visual display back to the data
Example: Over the course of a year, how does the daily number of births vary?
Example: Over the course of a year, how does the daily number of births vary?
- What patterns do you see?
Example
# Example codeggplot(data = ---, mapping = aes(---)) + geom_---(---) + coord_---() + scale_---_---() + ---# Create plotggplot(data = Births2015, mapping = aes(x = date, y = births)) + geom_point()
Example: Cycles related to day of the week?
# Create plotggplot(data = Births2015, mapping = aes(x = date, y = births, color = day_of_week)) + geom_point()- Additional aesthetic:
- Why is this not quite what we want? What do we want?
- What happened to the aspect ratio when we added the legend?
Example: Cycles related to day of the week?
# Create plotggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_point() + theme(legend.position = "bottom")You can also adjust the aspect ratio with the R chunk option:
fig.aspWhat if we want to see the direction that the number of births take over time for each day of the week?
- New visual cue/
geom?
- New visual cue/
Example
# Create plotggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_line() + theme(legend.position = "bottom")- What if we want visual cues for both position and direction?
Example
# Create plotggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_line() + geom_point() + theme(legend.position = "bottom")
Coordinate System Layer
library(lubridate)ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_line() + geom_point() + theme(legend.position = "bottom") + coord_cartesian(xlim = as_date(c("2015-01-01", "2015-01-31")))- How did this new layer change our plot?
Setting instead of Mapping
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = "midnightblue")) + geom_point()- What if we want all the points to be colored "midnightblue"?
Setting instead of Mapping
ggplot(data = Births2015, mapping = aes(x = date, y = births)) + geom_point(color = "midnightblue")- Here we'd say:
dateis mapped toxbirthsis mapped toy- "midnightblue" set the
color
Setting instead of Mapping
ggplot(data = Births2015, mapping = aes(x = date, y = births)) + geom_point(color = "#ff006e")- Can also supply hex codes for colors!
Setting instead of Mapping
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_point(color = "#ff006e") + geom_line() + theme(legend.position = "bottom")- Difference between setting/mapping aesthetics in the base layer or a specific
geomlayer
Layer order (sometimes) matters
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_line() + geom_point(color = "#ff006e") + theme(legend.position = "bottom")
Adding Curve(s)
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_point() + geom_smooth(method = "lm", se = FALSE) + theme(legend.position = "bottom")- Does a multiple linear regression line(s) capture the trend?
Adding Curve(s)
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_point() + geom_smooth(se = FALSE) + theme(legend.position = "bottom")- The default LOESS smoother usually does a reasonable job.
Adding Curve(s)
ggplot(data = Births2015, mapping = aes(x = date, y = births, color = wday)) + geom_point() + geom_smooth(color = "black", se = FALSE) + theme(legend.position = "bottom")- What happened?
Amounts: geom_bar
ggplot(data = movies, mapping = aes(x = binary)) + geom_bar()- Verbalize the mapping of data to geom_bar().
- How is this mapping different from geom_point()?
Another option: geom_col
# First wrangle with dplyrmovies_ag <- count(movies, binary) movies_ag## # A tibble: 2 × 2## binary n## <chr> <int>## 1 FAIL 863## 2 PASS 686ggplot(data = movies_ag, mapping = aes(x = binary, y = n)) + geom_col()
geom_point again
ggplot(data = movies_ag, mapping = aes(x = binary, y = n)) + geom_point(size = 4)- If you are worried about the data-ink ratio...
geom_point + geom_segment
ggplot(data = movies_ag, mapping = aes(x = binary, y = n)) + geom_segment(mapping = aes(xend = binary), yend = 0) + geom_point(size = 10, color = "orange") + ylim(c(0, 875))- Lollipop chart: compromise?
Two categorical variables: geom_bar
ggplot(data = movies, mapping = aes(x = rated, fill = binary)) + geom_bar()- Describe the mapping.
Two categorical variables: geom_bar
ggplot(data = movies, mapping = aes(x = rated, fill = binary)) + geom_bar(position = "fill")- Describe the mapping.
Two categorical variables: geom_bar
ggplot(data = movies, mapping = aes(x = rated, fill = binary)) + geom_bar(position = "dodge")- Describe the mapping.
Two categorical variables: geom_tile
movies_ag <- count(movies, rated, binary)ggplot(data = movies_ag, mapping = aes(x = rated, y = binary, fill = n)) + geom_tile()- Describe the mapping.
Two categorical variables: geom_tile
movies_ag <- count(movies, rated, binary)ggplot(data = movies_ag, mapping = aes(x = rated, y = binary, fill = n)) + geom_tile() + scale_fill_viridis_c(direction = -1)- Change the
fillscale!
Can display more than frequencies!
movies_ag <- group_by(movies, rated, binary) %>% summarize(mean_budget = mean(budget))ggplot(data = movies_ag, mapping = aes(x = rated, y = binary, fill = mean_budget)) + geom_tile() + scale_fill_viridis_c(direction = -1) + theme(legend.position = "bottom")
Distributions: geom_histogram
ggplot(movies, aes(x = budget)) + geom_histogram()- Verbalize the mapping.
Distributions: geom_histogram
ggplot(movies, aes(x = budget)) + geom_histogram(bins = 50, color = "white", fill = "darkcyan")- Can modify the mapping via the
binwidthorbinsarguments
Distributions: geom_histogram
ggplot(movies, aes(x = budget, fill = binary)) + geom_histogram(bins = 50, color = "white")- What is problematic about this graph?
Distributions: geom_histogram
ggplot(movies, aes(x = budget, fill = binary)) + geom_histogram(bins = 50, alpha = 0.4, position = "identity")- Still problematic
One option: Facetting
ggplot(movies, aes(x = budget, fill = binary)) + geom_histogram(bins = 50) + facet_wrap(~binary) + guides(fill = "none")
One option: Facetting
ggplot(movies, aes(x = budget, fill = binary)) + geom_histogram() + facet_grid(rated ~ binary, scales = "free_y") + guides(fill = "none")
Another option: geom_density
ggplot(movies, aes(x = budget, fill = binary)) + geom_density(alpha = 0.4) + theme(legend.position = "bottom")
Distributions: geom_density
ggplot(movies, aes(x = budget, fill = binary)) + geom_density(position = "fill") + theme(legend.position = "bottom")- How did the mapping change? What does this add to the story?
Distributions: geom_boxplot
ggplot(movies, aes(x = binary, y = budget)) + geom_boxplot()
Distributions: geom_boxplot
ggplot(movies, aes(x = binary, y = budget)) + geom_boxplot(varwidth = TRUE, notch = TRUE)- What does
varwidthdo? - Why might we add
notch = TRUE?
Distributions: geom_boxplot
ggplot(movies, aes(x = binary, y = budget, fill = rated)) + geom_boxplot()
Distributions: geom_violin
ggplot(movies, aes(x = binary, y = budget)) + geom_violin()- Utility of the violin over the box?
Distributions: geom_violin
ggplot(movies, aes(x = binary, y = budget)) + geom_violin() + geom_jitter(alpha = .1, width = .1, color = "darkcyan")