Visualizing Many Distributions at Once
Uploaded by
rohankumar809280Visualizing Many Distributions at Once
Uploaded by
rohankumar809280Visualizing Many Distributions at Once
To reveal patterns, variability, and relationships when dealing with
multiple groups of data
Compare multiple distributions simultaneously
Earlier methods like histograms or single distribution plots don’t scale well to many
groups
Example: weather data → showing temperature distributions for each of the 12
months in a year.
For that e use specialized visualization methods such as:
Boxplots
Violin plots
Ridgeline plots
Strip charts
Sina plots
Compare multiple distributions simultaneously
Key Terms
Response variable: The variable whose distribution is being studied (e.g.,
temperature).
Grouping variable: Defines subsets of data for comparison (e.g., months).
General approach: One axis shows the response variable, the other the grouping
variable.
Visualizing Distributions Along the Vertical Axis
Mean/Median with Error
Bars
Simplest approach: plot a point
(mean/median) with error
bars.
Problems:
Oversimplification → hides
distribution details.
Ambiguity → unclear whether it’s
mean/median.
Error bars’ meaning is
inconsistent (standard deviation?
standard error? CI?).
Misleading for skewed data.
Visualizing Distributions Along the Vertical Axis
Boxplots
Components
Median: central line.
Interquartile Range (IQR): 25th to
75th percentile.
Whiskers: extend up to 1.5×IQR from
box.
Outliers: dots beyond whiskers.
Advantages:
Compact → works well when comparing
many groups.
Shows skewness and spread.
Standardized → universally understood.
Visualizing Distributions Along the Vertical Axis
Boxplots
• Work well when plotted next to
each other to visualize many
distributions at once.
• Reveals December temperatures
are skewed (long tail for extremely
cold days)
• Limitation: Cannot reveal
multimodality (two peaks).
Visualizing Distributions Along the Vertical Axis
Violin Plots
Violin plots extend boxplots by
showing distribution shape
through kernel density estimation
(KDE).
Width represents density at that
value.
Symmetric shape → mirrored KDE.
Can reveal multiple peaks
(multimodality).
Visualizing Distributions Along the Vertical Axis
When to use:
Large sample sizes.
Detecting subtle distribution
shapes.
Limitations:
Requires enough data → small
samples make density misleading.
May suggest smooth patterns that
don’t exist.
Visualizing Distributions Along the Vertical Axis
Lincoln temperatures as violin plots.
Reveals bimodality in November
(two peaks at ~35°F & ~50°F).
Boxplot cannot reveal this.
More detailed than boxplots.
Needs large sample sizes, otherwise
smoothing creates misleading shapes.
Visualizing Distributions Along the Vertical Axis
Strip Charts
Definition: Plot all data points
individually (raw visualization).
Problem: Overplotting (points
overlap, hiding density).
Visualizing Distributions Along the Vertical Axis
Jittering
Add random horizontal noise
so points don’t overlap.
Reveals frequency without
density smoothing.
Best for small to medium
samples.
Not suitable for large datasets
(becomes unreadable).
Visualizing Distributions Along the Vertical Axis
Sina Plots
Sina plot of Lincoln temperatures.
Hybrid of violin plot + jittered points.
Points spread horizontally
proportional to density at that value.
Shows both individual
observations and overall
distribution.
Best for medium datasets.
Combines advantages of violins
(shape) and strip charts (raw points).
Visualizing distributions along the horizontal axis
Ridgeline Plots
Each month’s distribution stacked
vertically.
November’s two clusters (35°F,
50°F) clearly visible.
Looks like mountain ridges → hence
the name.
Excellent for time series
comparisons.
Very intuitive for showing gradual
changes.
Visualizing distributions along the horizontal axis
Scaling to Large Datasets
Movie lengths (1913–2005) as
ridgeline plot.
Nearly 100 distributions
shown.
Early decades = wide variety of
lengths.
Since 1960s = standardization
at ~90 minutes.
Ridgeline plots scale well to
large numbers of groups.
Visualizing distributions along the horizontal axis
Comparing Two Groups
Over Time
Voting patterns in U.S. House
of Representatives (1963–
2013).
Uses DW-NOMINATE
scores to represent ideology.
Two ridgeline distributions at
each Congress session:
Democrats (blue), Republicans
(red).
Shows increasing
polarization over time.
Perfect for showing divergence
between two categories across
time.
Thank You!
