Data Visualization

Introduction to Graphic Representation of

Data
• Data visualization is the process of representing data through
charts, graphs, and other visual formats.
• It makes complex data more accessible, understandable, and
usable.
• Different types of charts are suitable for different data types:
single variable, two variables, and more than two variables.
Characteristics of Effective Graphical
Displays
• Clarity: The graph should be easy to understand, with clear labels,
scales, and legends.
• Accuracy: The graph must accurately represent the data without
distortion or misleading interpretations.
• Simplicity: Avoid unnecessary decorations or complexity that may
obscure the data’s meaning.
• Relevance: The chosen graph type should be appropriate for the data
being displayed.
• Consistency: Use consistent scales and labels, especially when
comparing multiple graphs.
• Visual Appeal: Aesthetic design should not overshadow clarity but
enhance the viewer's ability to interpret the data.
Single variable (Univariate)
• Single variable plots, also known as univariate plots, are used to
visualize the distribution and characteristics of a single variable
within a dataset.
• These plots are essential in understanding the central tendency,
spread, and overall shape of the data.
Chart Types :
Charts for Single Variables (Univariate Data)
• Dot Plot: Displays data points as dots; useful for small data sets.
• Example: Distribution of test scores in a class.
• Jitter Plot: Similar to a dot plot but with added random noise to
spread out overlapping points.
• Example: Scores of multiple tests with slight overlaps.
• Pie Chart: Shows proportions of categories as parts of a circle.
• Example: Market share of companies in a sector.
• Note: Use pie charts sparingly; they are less effective for precise
comparisons.
• Box-and-Whisker Plot: Shows distribution using quartiles;
highlights outliers.
• Example: Income distribution in a region.
• Histogram: Represents the frequency of data within intervals
(bins).
• Example: Age distribution of people in a city.
1. Dot Plot
• Description: A dot plot displays individual data points along a single
axis. Each dot represents one observation, and dots may be stacked to
represent frequency.
• When to Use: Dot plots are useful for small to medium-sized datasets
to visualize the frequency and distribution of data points. They are
especially helpful when you want to see individual values and their
distribution.
• Pros:
• Shows individual data points.
• Easy to interpret for small datasets.
• Cons:
• Becomes cluttered with large datasets.
• Example Use Case: Displaying the number of students scoring each
grade in a small class.
2. Jitter Plot
• Description: A jitter plot is similar to a dot plot but with random noise
added to prevent overlapping of data points (jittering). This makes it
easier to see individual points when many values are the same.
• When to Use: Jitter plots are useful when you have a large number of
identical data points that would overlap in a standard dot plot. The jitter
helps to spread out the points and avoid overplotting.
• Pros:
• Prevents overplotting.
• Useful for large datasets with many identical values.
• Cons:
• The added noise can slightly distort the perception of data distribution.
• Example Use Case: Visualizing the distribution of exam scores where
many students have the same scores.
3. Error Bar Plot
• Description: An error bar plot shows the mean or median of data points
along with error bars that represent variability (e.g., standard deviation,
standard error).
• When to Use: Error bar plots are used when you want to show not only the
central tendency of the data but also the variability or uncertainty around it.
They are common in scientific and experimental data analysis.
• Pros:
• Provides insight into data variability and reliability.
• Useful for comparing the central tendency and spread across different groups.
• Cons: Interpretation can be complex, especially with overlapping error bars.
• Example Use Case: Comparing the average response time in different
experimental conditions with error bars representing standard deviation.
• x and y:
• x: The data points for the x-axis.
• y: The data points for the y-axis.
• These are required parameters and represent the data being
plotted.
• yerr and xerr:
• yerr: The error values in the y-direction.
• xerr: The error values in the x-direction.
4. Box-and-Whisker Plot (Box Plot)
• Description: A box plot displays the distribution of data based on five
summary statistics: minimum, first quartile (Q1), median, third quartile (Q3),
and maximum. The box represents the interquartile range (IQR), and the
"whiskers" extend to the smallest and largest values within 1.5 times the IQR.
• When to Use: Box plots are ideal for comparing distributions between groups
or identifying outliers. They provide a concise summary of data spread and
central tendency.
• Pros:
• Summarizes data distribution.
• Identifies outliers.
• Useful for comparing multiple groups.
• Cons:
• Does not show individual data points.
• Example Use Case: Comparing the distribution of salaries across different
departments in a company.
5. Histogram
• Description: A histogram divides the data into bins (intervals) and counts the
number of observations in each bin. It displays the frequency distribution of a
continuous variable.
• When to Use: Histograms are used to visualize the shape of the distribution
of a dataset, such as whether it is normal, skewed, or multimodal. They are
particularly useful for large datasets.
• Pros:
• Shows the shape of the distribution.
• Can handle large datasets.
• Cons:
• Choice of bin width can affect the interpretation.
• Does not show individual data points.
• Example Use Case: Displaying the distribution of household incomes in a
city.
Choosing the Right Plot
• Small Datasets:
• Dot Plot: Use when you want to visualize individual data points and their frequencies.
• Jitter Plot: Use when there is significant overlap in data points in a dot plot.
• Data with Variability Information:
• Error Bar Plot: Use when you want to convey the central tendency and variability or
uncertainty in the data.
• Comparing Distributions:
• Box-and-Whisker Plot: Use for summarizing and comparing distributions across
multiple groups, while also identifying outliers.
• Visualizing Distribution Shape:
• Histogram: Use when you want to understand the overall shape, skewness, and
modality of the data distribution.
• Each of these plots serves a specific purpose and helps in understanding
different aspects of the data. The choice of plot depends on the dataset size,
the type of data (categorical or continuous), and the specific insights you aim
to gain.
Two-variable plots
• Two-variable plots, also known as bivariate plots, are used to
explore the relationship between two variables.
• These plots help in identifying patterns, correlations, trends, and
potential anomalies between the variables.
• Below is a detailed explanation of various two-variable plots,
when to use them, and which graphs are most suitable for
different types of data.
1. Bar Chart
• Description: A bar chart displays data using rectangular bars where
the length of the bar is proportional to the value of the variable. Bar
charts can compare two variables by using grouped or stacked bars.
• When to Use: Bar charts are used when comparing categorical data
between two groups. They are particularly useful when you want to
show differences in magnitude between categories.
• Pros:
• Easy to interpret.
• Effective for comparing discrete categories.
• Cons: Not suitable for continuous data.
• Example Use Case: Comparing the sales figures of different products
across two different years.
2. Scatter Plot
• Description: A scatter plot displays individual data points plotted on a
two-dimensional graph with one variable on the x-axis and the other on
the y-axis. It shows how one variable is related to another.
• When to Use: Scatter plots are ideal for examining the relationship or
correlation between two continuous variables. They are useful for
detecting patterns, trends, clusters, and outliers.
• Pros:
• Visualizes relationships and correlations.
• Identifies patterns, clusters, and outliers.
• Cons:
• Can be cluttered with large datasets.
Example Use Case: Exploring the relationship between advertising
spend and sales revenue.
3. Line Plot
• Description: A line plot connects individual data points with lines,
typically used to display trends over time. Each data point represents
the value of a variable at a specific time or ordered sequence.
• When to Use: Line plots are used for visualizing trends in time series
data or any sequential data. They are particularly useful for showing
changes and trends over time.
• Pros:
• Effective for visualizing trends over time.
• Clear representation of changes in data.
• Cons: Not suitable for non-sequential data.
• Example Use Case: Tracking the monthly temperature change over a
year.
4. Log-Log Plot
• Description: A log-log plot is a scatter plot where both the x-axis and y-
axis are on a logarithmic scale. This type of plot is used when the data
spans several orders of magnitude.
• When to Use: Log-log plots are used when both variables have a
multiplicative relationship, or when dealing with data that spans
multiple orders of magnitude. They are commonly used in scientific
data analysis.
• Pros:
• Useful for visualizing power-law relationships.
• Handles wide-ranging data scales.
• Cons:
• Can be difficult to interpret without proper knowledge.
• Example Use Case: Analyzing the relationship between the size of an
earthquake and the energy released.
Choosing the Right Plot
• Comparing Categorical Data:
• Bar Chart: Use when you need to compare categorical variables across different
groups. For example, comparing test scores between two different classes.
• Exploring Relationships Between Continuous Variables:
• Scatter Plot: Ideal for visualizing the correlation or relationship between two
continuous variables, such as height and weight of individuals.
• Visualizing Trends Over Time:
• Line Plot: Best for showing trends and changes over time, such as stock prices
over months or years.
• Analyzing Data with Wide Ranges:
• Log-Log Plot: Use when your data spans several orders of magnitude and you
suspect a multiplicative relationship between the variables.
Summary of When to Use Which Plot:
• Bar Chart: Compare categories between two variables (especially
categorical variables).
• Scatter Plot: Explore relationships and correlations between two
continuous variables.
• Line Plot: Track changes or trends over time or ordered sequences.
• Log-Log Plot: Analyze data that spans several orders of magnitude with
multiplicative relationships.
• These two-variable plots are fundamental tools in data analysis,
helping to uncover insights about relationships, trends, and patterns
between variables. The choice of plot depends on the nature of the
data and the specific analytical goals.
More than two-variable plots
• More than two-variable plots, also known as multivariate plots,
are used to visualize relationships between three or more
variables in a dataset.
• These plots help in understanding complex interactions and
patterns that cannot be captured by two-variable plots.
• Below is a detailed explanation of various multivariate plots, when
to use them, and which graphs are most suitable for different
types of data.
1. Stacked Plot
• Description: A stacked plot visualizes the cumulative contribution of multiple
variables over a single dimension, often time. Each segment (or "stack") of the plot
represents one variable's contribution, with the segments stacked on top of each
other.
• When to Use: Stacked plots are used to show how different variables contribute to
a total over time or another dimension. They are useful for understanding the
proportion of each variable relative to the whole.
• Pros:
• Shows the composition of multiple variables.
• Useful for displaying cumulative totals.
• Cons:
• Can be difficult to interpret if there are too many variables.
• Changes in individual variables are harder to track.
• Example Use Case: Visualizing the sales contribution of different product
categories over several months.
2. Parallel Coordinate Plot
• Description: A parallel coordinate plot visualizes multiple variables by plotting each
variable on a separate vertical axis. Each data point is represented as a line
connecting the axes.
• When to Use: Parallel coordinate plots are used when you need to compare many
variables across different observations simultaneously. They are particularly useful
for identifying patterns, correlations, and outliers in high-dimensional data.
• Pros:
• Handles high-dimensional data.
• Good for comparing multiple variables simultaneously.
• Cons:
• Can become cluttered with large datasets.
• Interpretation can be challenging without careful design.
• Example Use Case: Analyzing the characteristics of different types of cars, such as
engine size, fuel efficiency, and price.
3. Scatter Matrix (Pair Plot)
• Description: A scatter matrix is a grid of scatter plots for each pair of
variables in a dataset. Each cell in the grid shows the relationship between
two variables, with histograms along the diagonal to show the distribution of
individual variables.
• When to Use: Scatter matrices are useful when you want to explore the
relationships between all pairs of variables in a dataset. They are particularly
helpful in identifying correlations, patterns, and potential multicollinearity.
• Pros:
• Provides a comprehensive view of relationships between multiple variables.
• Helps in identifying correlations and patterns.
• Cons:
• Becomes overwhelming with too many variables.
• Can be hard to interpret without careful analysis.
• Example Use Case: Exploring relationships between various financial
indicators such as revenue, profit, expenses, and stock price.
4. Heatmap
• Description: A heatmap is a two-dimensional representation of data where
the individual values are represented by colors. It is commonly used to
visualize correlations between multiple variables in a matrix format.
• When to Use: Heatmaps are ideal for visualizing the correlation or
relationship between many variables in a compact and intuitive manner. They
are particularly useful in identifying clusters, patterns, and correlations.
• Pros:
• Compact and easy to interpret.
• Effective for visualizing large datasets.
• Cons:
• May oversimplify complex relationships.
• Choice of color scale can impact interpretation.
• Example Use Case: Displaying the correlation matrix of various economic
indicators like GDP, inflation rate, unemployment rate, and interest rate.
Choosing the Right Plot
• Understanding Composition and Contribution:
• Stacked Plot: Use when you want to show how different variables
contribute to a total over a single dimension, such as time.
• Comparing Multiple Variables:
• Parallel Coordinate Plot: Ideal for comparing multiple variables across
different observations, especially in high-dimensional datasets.
• Scatter Matrix: Use when you want to explore pairwise relationships
between all variables in a dataset.
• Visualizing Relationships and Correlations:
• Heatmap: Best for visualizing correlations between many variables in a
compact and intuitive format.
Summary of When to Use Which Plot:
• Stacked Plot: Use for cumulative contributions over time or
another dimension.
• Parallel Coordinate Plot: Best for high-dimensional data
comparisons.
• Scatter Matrix: Useful for pairwise exploration of relationships.
• Heatmap: Ideal for visualizing correlations.
• These multivariate plots are essential tools in understanding
complex datasets with multiple variables. The choice of plot
depends on the nature of the data, the specific insights you aim to
gain, and the ease of interpretation required for your analysis.
Conclusion:
• Choosing the right chart type is essential for effective data
communication.
• Consider the nature of the data (univariate, bivariate, multivariate)
and the message you wish to convey.
• Following best practices ensures that your visualizations are clear,
accurate, and engaging.
• Refer notebook :Visualization_plots for all plots.
 Data and Data Visualization
• 1. Data and Data Taxonomy:
• Data can be classified based on its dimensionality, structure, and type:
• 1D Data: Data consisting of a single attribute or variable, e.g., height, weight, income.

• Temporal Data: Data indexed by time, e.g., stock prices, weather data, sensor logs.

• 2D Data: Data with two attributes, often represented as pairs (x, y), e.g., geographic coordinates, scatterplots.

• 3D Data: Data with three attributes, e.g., x, y, z in 3D models or volumetric measurements.

• 4D Data: Extends to three spatial dimensions plus an additional dimension (often time), e.g., climate simulations over time.

• ND Data (High-dimensional): Data with more than 4 attributes, common in machine learning and genomics.
• Trees: Hierarchical data structure where nodes have parent-child relationships, e.g., file systems, organizational charts.

• Graphs: Data represented as nodes and edges, suitable for relationships and networks, e.g., social networks, transport
systems.
• Types of Data (by Measurement Scale):
• Nominal Data: Categories without order, e.g., colors, gender.
• Ordinal Data: Ordered categories, e.g., ranks, Likert scales.
• Suppose in a race, participants are ranked 1st, 2nd, 3rd=
• The order is clear → 1st performed better than 2nd, 2nd better than
3rd.
• But the gap between positions is not known → The time difference
between 1st and 2nd may be 0.1 sec, while between 2nd and 3rd it
may be 5 sec.
• Hence, ranks show order but not exact magnitude of difference.
• A Likert scale is commonly used in questionnaires:

• Example statement: “I am satisfied with this service.”

• Responses:

• Strongly Disagree

• Disagree

• Neutral

• Agree

• Strongly Agree

• Here, the order is meaningful: Strongly Disagree < Disagree < Neutral < Agree < Strongly Agree.

• But the intervals are subjective: the gap between “Disagree” and “Neutral” may not be the same as between “Neutral” and
“Agree.”
• Quantitative Data: Numerical values, subdivided into:
• Quantitative data deals with numbers and allows us to perform arithmetic operations. It is subdivided
into two main types:
• 1. Interval Data
• Definition: Numerical data where the differences between values are meaningful, but there is no true
zero point.
• You can add and subtract values, but ratios (multiplication/division) don’t make sense.
• Example: Temperature in Celsius (or Fahrenheit)
• 20°C is 10 degrees warmer than 10°C → meaningful difference
• But 20°C is not twice as hot as 10°C because 0°C does not mean “no temperature.” It’s just an arbitrary
point on the scale.
• Another example: Calendar years (2000, 2025). The difference (25 years) is meaningful, but “Year 0” is
not a true absence of time.

• Interval: Meaningful differences but no true zero, e.g., temperature in Celsius.

• Ratio Data
• Definition: Numerical data with a true zero that represents the absence of the quantity. Both
differences and ratios are meaningful.
• Key property: You can add, subtract, multiply, and divide values.
• Example: Weight
• 60 kg is 20 kg heavier than 40 kg → difference is meaningful
• 60 kg is also 1.5 times heavier than 40 kg → ratio is meaningful
• Here, 0 kg means complete absence of weight (true zero).
• Other examples:
• Height (0 cm = no height).
• Income (₹0 = no money).
• Distance (0 km = no distance).
• Time duration (0 sec = no time).
Hence :
• Quantitative Data: Numerical values, subdivided into:

1. Interval: Meaningful differences but no true zero, e.g.,

temperature in Celsius.
2. Ratio: Has a true zero, e.g., weight, height.
Visualization
• Visualization is the graphical representation of data to aid
understanding, analysis, and communication.
• Importance of Data Visualization:
• Transforms complex data into interpretable insights.
• Identifies patterns, trends, and outliers.
• Enhances storytelling and decision-making.
• Bridges the gap between data producers and consumers.
Goals of Visualization:
• Exploration: Understand and discover hidden structures in raw
data.
• Explanation: Communicate findings clearly to others.
• Confirmation: Validate hypotheses and test models.
• Presentation: Convey data-driven narratives to broader
audiences.
Characteristics of Good Visualizations:
• Clarity: Easy to interpret without unnecessary complexity.
• Accuracy: Truthful representation of data.
• Efficiency: Conveys information quickly and effectively.
• Aesthetics: Appealing visuals that engage without distracting.
• Relevance: Matches visualization method to the nature of the
data.
Exploratory Data Analysis (EDA)
• EDA is the process of summarizing main characteristics of data
often with visual methods:
• Helps detect errors, missing values, and anomalies.
• Reveals distributions, correlations, and structures.
• Uses tools like histograms, scatterplots, boxplots, and heatmaps.
• Supports hypothesis generation before formal modeling.
Time Series Data and Visualization:
• Time series data shows how variables change over time.
• Time series data is a collection of observations recorded at successive
time intervals (daily, monthly, yearly, etc.).
• It captures how variables change over time.
• Common in economics, finance, weather, sales, biology, and
engineering.
• Examples of Time Series Data:
• Stock Market Prices – Daily closing prices of a company’s stock.
• Weather Data – Daily temperature, rainfall, humidity over a year.
• Sales Data – Monthly revenue of a retail store.
• Sensor Data – IoT device recording heart rate every second.
Visualization of Time Series Data
• Time series visualization helps to detect trends, seasonality, and anomalies.
• 1. Line Graph
• Most common visualization.
• Time is placed on the x-axis, variable values on the y-axis.
• Example: Plotting daily COVID-19 cases over a year.

• 2. Bar Chart (Time-based)

• Bars represent values at discrete time intervals (e.g., months, quarters).
• Example: Annual sales revenue comparison.

• [Link]
• Displays intensity of data across time and categories.
• Example: Daily website traffic over hours (rows = days, columns = hours).

• 5. Multiple Time Series (Comparative Line Plots)

• Comparing several variables over time.
• Example: Plotting rainfall and temperature trends together.
Text and Visualization:
• Text data is unstructured and requires preprocessing.

• Common visualizations: word clouds, co-occurrence networks,

topic modeling maps, sentiment heatmaps.

• Helps in understanding themes, trends, and emotional tone of

textual content.
Tools for Visualization
• Popular tools and libraries for creating visualizations:

• Python Libraries: Matplotlib, Seaborn, Plotly, Bokeh, Altair.

• R Libraries: ggplot2, lattice, plotly in R.

• General Tools: Tableau, Power BI, [Link] (JavaScript), Excel.

• Emerging Tools: Observable, Flourish, RAWGraphs.

• Each tool varies in complexity, flexibility, and purpose:

• Tableau/Power BI: User-friendly dashboards for business intelligence.

• Matplotlib/Seaborn/ggplot2: Statistical and scientific visualization.

Unit 2: Collecting, Pre-processing, and
Visualizing Data
• 1. Data Collection
• Data collection is the first step in data analysis and visualization. Sources can vary
widely depending on the domain and problem.
• Single Source: Data obtained from a single dataset or source.
• Example: Sales data from a company’s database.
• Advantages: Easier integration and cleaning.
• Limitations: May lack context and completeness.

• Multiple Sources: Data combined from different datasets or repositories.

• Example: Merging weather data with crop yield data for agricultural research.
• Advantages: Richer, more comprehensive insights.
• Challenges: Data integration, matching formats, handling inconsistencies.
• Web Scraping: Automated extraction of data from websites.
• Tools: Python (BeautifulSoup, Selenium, Scrapy).
• Applications: Collecting product data from e-commerce sites,
extracting articles from news portals, gathering reviews and
ratings.
• Considerations: Ethical/legal boundaries, [Link], server load,
dynamic content handling.
2. Data Pre-processing
• Pre-processing involves preparing raw data for analysis and visualization.

• Data Cleaning: Removing noise and errors.

• Handling missing values (drop, impute).

• Removing duplicates.
• Correcting inconsistent formats (dates, currency, units).

• Data Aggregation: Combining data at different levels.

• Summarization: e.g., daily to monthly sales totals.

• Grouping: e.g., average test scores per department.
• Benefits: Reduces complexity, makes patterns visible.
3. Mapping Data onto Aesthetics
• Visualization involves mapping data values onto visual properties (aesthetics) like position, shape,
color, or size.

• Aesthetics and Types of Data:

• Position: Used for quantitative and ordinal data.

• Color: Encodes categorical differences (nominal/ordinal data) or gradients for continuous data.
• Size: Indicates magnitude in quantitative data.
• Shape: Differentiates categories (nominal data).
• Line/Area: Represents changes over time or aggregation.

• Scales map data values onto aesthetics:

• A scale defines how data values are converted into visual variables.
• Example: A continuous scale maps numerical values onto a color gradient; a categorical scale
assigns distinct colors to classes.
• Good use of scales ensures clarity and avoids misinterpretation.
4. Coordinate Systems and Axes
• Coordinate systems provide the framework for plotting data.
• Cartesian Coordinates:
• Most common system (x, y axes at right angles).
• Suitable for scatterplots, bar charts, and line charts.
• Intuitive for representing quantitative and categorical data.

• Nonlinear Axes:
• Useful when data spans wide ranges or exponential growth.
• Examples: Logarithmic scales for population growth, Richter scale for earthquakes.
• Helps in compressing large ranges into readable visualizations.

• Coordinate Systems with Curved Axes:

• Examples: Polar coordinates, radial plots, geographic projections.
• Polar coordinates → used for pie charts, radar charts.
• Geographic coordinate systems → used in maps to represent latitude and longitude.
• Effective for data where relationships are cyclical (time of day, seasons) or spatial.