Why scatter plots show relationships in Matplotlib - Why It Works This Way

Overview - Why scatter plots show relationships

What is it?

A scatter plot is a simple graph that shows points representing two sets of data. Each point's position on the horizontal and vertical axes shows values from two variables. This helps us see if there is a pattern or connection between these variables. Scatter plots are often used to explore how one thing might affect or relate to another.

Why it matters

Without scatter plots, it would be hard to quickly see if two things are connected or not. For example, if you want to know if studying more hours leads to better test scores, scatter plots let you spot that pattern easily. They help people make decisions based on data by showing relationships visually instead of just numbers.

Where it fits

Before learning scatter plots, you should understand basic graphs like line and bar charts. After mastering scatter plots, you can explore more advanced topics like correlation, regression, and data modeling. Scatter plots are a foundation for understanding how variables interact in data science.

Mental Model

Core Idea

A scatter plot shows how two variables move together by placing points on a grid where each point's position reflects both variables' values.

Think of it like...

Imagine you have a map where each point shows a friend's location based on how far east and north they are. The pattern of points tells you if friends tend to gather in certain areas or spread out randomly.

  Y-axis (Variable 2)
    ↑
    │    ●       ●
    │       ●
    │  ●
    │          ●
    │●
    └────────────────→ X-axis (Variable 1)
Each dot's position shows one pair of values from two variables.

Build-Up - 7 Steps

1

Foundation

Understanding variables and data points

Concept: Learn what variables are and how data points represent pairs of values.

Variables are things we measure or observe, like height or age. Each data point in a scatter plot shows one pair of values, one from each variable. For example, a point might show a person's height on the X-axis and their weight on the Y-axis.

Result

You can identify what each point means and how it relates to the two variables.

Understanding that each point represents a pair of values is the base for seeing relationships between variables.

2

Foundation

Plotting points on two axes

Concept: Learn how to place points on a graph using two axes for two variables.

The horizontal axis (X-axis) shows values of one variable, and the vertical axis (Y-axis) shows values of the other. To plot a point, find the X value on the bottom and the Y value on the side, then mark where they meet.

Result

You get a visual map of all data points showing how values pair up.

Knowing how to plot points correctly is essential to create meaningful scatter plots.

3

Intermediate

Identifying patterns in scatter plots

🤔Before reading on: do you think points clustered together always mean a strong relationship? Commit to your answer.

Concept: Learn to spot common patterns like clusters, trends, or randomness in scatter plots.

When points form a line or curve, it suggests a relationship between variables. If points are scattered randomly, there may be no connection. Clusters show groups with similar values. For example, points rising from left to right show a positive trend.

Result

You can visually guess if variables are related and how.

Recognizing patterns helps you quickly understand the nature of relationships in data.

4

Intermediate

Using scatter plots to detect correlation

🤔Before reading on: do you think a perfect straight line is the only sign of correlation? Commit to your answer.

Concept: Understand how scatter plots reveal correlation strength and direction between variables.

Correlation means how strongly two variables move together. A tight line upward means strong positive correlation; downward means strong negative correlation. A fuzzy cloud means weak or no correlation. Scatter plots let you see this visually before calculating numbers.

Result

You can estimate correlation by looking at the plot shape.

Seeing correlation visually is faster and often more intuitive than just numbers.

5

Intermediate

Adding color and size for extra data

Concept: Learn how to use point color and size to show more information in scatter plots.

Besides position, points can have colors or sizes representing other variables. For example, color might show categories like gender, and size might show income level. This adds layers of insight in one plot.

Result

You get a richer, multi-dimensional view of data relationships.

Using visual cues beyond position helps reveal complex patterns in data.

6

Advanced

Scatter plots in regression analysis

🤔Before reading on: do you think scatter plots only show relationships but cannot help predict values? Commit to your answer.

Concept: Learn how scatter plots support fitting lines to predict one variable from another.

Regression fits a line or curve through points to model relationships. Scatter plots show the data points and the fitted line, helping check how well the model fits. This is key for predictions and understanding cause-effect.

Result

You can visually assess model accuracy and relationship strength.

Scatter plots are not just for seeing data but also for building and validating predictive models.

7

Expert

Limitations and pitfalls of scatter plots

🤔Before reading on: do you think scatter plots always reveal true relationships without bias? Commit to your answer.

Concept: Understand when scatter plots can mislead due to outliers, scale, or hidden variables.

Outliers can distort patterns, and axis scales can exaggerate or hide trends. Also, scatter plots show correlation but not causation. Hidden third variables might cause apparent relationships. Experts use scatter plots carefully with these limits in mind.

Result

You avoid wrong conclusions from misleading scatter plots.

Knowing scatter plot limits prevents common data interpretation errors.

Under the Hood

Scatter plots work by mapping each data pair to a coordinate system where the X and Y axes represent two variables. The plotting system translates data values into pixel positions on the screen. This visual mapping allows the human eye to detect patterns, clusters, and trends quickly, leveraging our natural spatial recognition.

Why designed this way?

Scatter plots were designed to provide a simple, intuitive way to visualize relationships between two variables without complex calculations. Early statisticians needed a tool to quickly spot trends and outliers in data. Alternatives like tables or raw numbers were harder to interpret visually, so scatter plots became a standard exploratory tool.

Data pairs → Coordinate mapping → Pixel positions

┌───────────────┐
│ Data pairs    │
│ (x, y values) │
└──────┬────────┘
       │
       ▼
┌───────────────┐
│ Coordinate    │
│ system maps   │
│ values to     │
│ positions     │
└──────┬────────┘
       │
       ▼
┌───────────────┐
│ Screen pixels │
│ plot points   │
└───────────────┘

Myth Busters - 4 Common Misconceptions

Quick: Does a cluster of points always mean a strong relationship? Commit yes or no.

Common Belief: If points cluster tightly, it means a strong relationship between variables.

Reality: Clusters can occur due to grouping or categories, not necessarily a direct relationship between variables.

Why it matters: Misinterpreting clusters can lead to false assumptions about how variables influence each other.

Quick: Is a straight line the only sign of correlation? Commit yes or no.

Common Belief: Only a straight line pattern shows correlation between variables.

Reality: Correlation can be nonlinear; curved patterns also show relationships but need different analysis.

Why it matters: Ignoring nonlinear relationships can miss important insights in data.

Quick: Does correlation in scatter plots prove one variable causes the other? Commit yes or no.

Common Belief: If two variables correlate in a scatter plot, one causes the other.

Reality: Correlation does not imply causation; other factors may cause the observed pattern.

Why it matters: Assuming causation leads to wrong decisions and flawed conclusions.

Quick: Can changing axis scales affect how relationships appear? Commit yes or no.

Common Belief: Axis scales do not affect the interpretation of scatter plots.

Reality: Changing scales can exaggerate or hide relationships, misleading interpretation.

Why it matters: Misreading plots due to scale manipulation can cause incorrect data analysis.

Expert Zone

1

Scatter plots can reveal heteroscedasticity, where the spread of points changes with variable values, important for regression assumptions.

2

Using transparency (alpha) in points helps visualize dense areas without overplotting, a subtle but powerful visualization technique.

3

Choosing axis limits carefully avoids misleading impressions of relationships, a detail often overlooked even by experienced analysts.

When NOT to use

Scatter plots are not suitable for categorical variables without numeric meaning or for very large datasets where overplotting hides patterns. Alternatives include box plots for categories or hexbin plots for large data.

Production Patterns

In real-world data science, scatter plots are used for exploratory data analysis, model diagnostics (residual plots), and communicating findings visually. They often combine with regression lines, confidence intervals, and interactive features in dashboards.

Connections

Correlation coefficient

Scatter plots visually build on the concept of correlation by showing data points whose pattern reflects correlation strength and direction.

Understanding scatter plots helps grasp what correlation numbers mean in real data.

Regression analysis

Scatter plots provide the foundation for regression by displaying data points that regression lines try to fit.

Seeing scatter plots clarifies how regression models relate variables and predict outcomes.

Geographic mapping

Scatter plots and maps both plot points on two-dimensional grids to reveal spatial patterns.

Recognizing this connection helps understand how visualizing data spatially uncovers hidden relationships.

Common Pitfalls

#1 Ignoring outliers that distort the pattern.

Wrong approach:

plt.scatter(x, y)
# No check for outliers or filtering

Correct approach:

filtered_x = [xi for xi, yi in zip(x, y) if is_not_outlier(xi, yi)]
filtered_y = [yi for xi, yi in zip(x, y) if is_not_outlier(xi, yi)]
plt.scatter(filtered_x, filtered_y)

Root cause:

Not recognizing that extreme points can mislead the visual interpretation.

#2 Using inappropriate axis scales that exaggerate trends.

Wrong approach:

plt.scatter(x, y)
plt.xlim(0, 10)
plt.ylim(0, 1000) # Very different scales

Correct approach:

plt.scatter(x, y)
plt.xlim(0, 1000)
plt.ylim(0, 1000) # Matching scales for fair view

Root cause:

Lack of attention to axis scaling causing visual distortion.

#3 Plotting categorical variables as numeric without encoding.

Wrong approach:

plt.scatter(['red', 'blue', 'green'], [1, 2, 3])

Correct approach:

colors_encoded = {'red':1, 'blue':2, 'green':3}
plt.scatter([colors_encoded[c] for c in ['red', 'blue', 'green']], [1, 2, 3])

Root cause:

Misunderstanding that scatter plots require numeric axes.

Key Takeaways

Scatter plots visually map pairs of variable values as points on two axes, revealing relationships.

Patterns like lines, clusters, or randomness in scatter plots indicate different types of relationships or lack thereof.

Scatter plots help estimate correlation strength and direction before formal calculations.

They support advanced analysis like regression by showing data and fitted models together.

Understanding scatter plot limitations prevents misinterpretation and poor decisions based on data.