Data Science Data Visualization Syllabus

Why is encoding quantitative data by position more accurate than by area or color?

Human visual perception judges position along a common scale most precisely, whereas area and color are judged with much greater error, making them suited only to approximate or categorical comparisons.

What is a colorblind-safe palette and why does it matter?

A palette (e.g., avoiding red-green contrasts, or using ColorBrewer/viridis) chosen so that ~8% of men and ~0.5% of women with color vision deficiency can still distinguish encoded categories.

What is the difference between sequential, diverging, and qualitative color scales?

Sequential encodes ordered low-to-high values (single hue ramp); diverging encodes deviation from a midpoint (two hues meeting at neutral); qualitative encodes unordered categories (distinct hues).

Name three programming languages commonly used for data visualization.

Python, R, and JavaScript (with SQL often used for data retrieval).

Which Python libraries are most commonly used for visualization?

Matplotlib, Seaborn, Plotly, Bokeh, and Altair.

What is the primary visualization package in R, and what grammar does it implement?

ggplot2, which implements the Grammar of Graphics.

What is the Grammar of Graphics?

A framework (by Leland Wilkinson) that builds visualizations from layered components: data, aesthetic mappings, geometric objects (geoms), scales, statistical transformations, coordinates, and facets.

Which JavaScript library gives the most low-level control for custom web visualizations?

D3.js (Data-Driven Documents), which binds data to DOM elements and manipulates them with web standards (SVG, HTML, CSS).

Name three end-user BI software platforms for visualization.

Tableau, Microsoft Power BI, Looker, Qlik, and Google Data Studio/Looker Studio (any three).

What is the difference between a database and a data warehouse as a data source?

A database (often OLTP) is optimized for transactional reads/writes of current data; a data warehouse (OLAP) is optimized for analytical querying of large, integrated, historical data for visualization and reporting.

What does ETL stand for in data sourcing for visualization?

Extract, Transform, Load — the process of pulling data from sources, cleaning/reshaping it, and loading it into a destination such as a warehouse for analysis.

What is the difference between structured, semi-structured, and unstructured data sources?

Structured = tabular with a fixed schema (SQL tables); semi-structured = tagged/self-describing (JSON, XML); unstructured = no predefined model (text, images, video).

When choosing a chart, what is the first question to ask?

What relationship or message you want to show — comparison, composition, distribution, or relationship — which then narrows the appropriate chart family.

Why is a pie chart often discouraged for more than a few categories?

Humans judge angles and areas poorly, so comparing many similarly-sized slices is inaccurate; a bar chart with position encoding is usually clearer.

Which chart should you use to compare values across many categories?

A bar chart (horizontal if category labels are long), since length/position along a common scale is read accurately.

What is data preprocessing in the visualization pipeline?

Cleaning and preparing raw data before plotting — handling missing values, removing duplicates, correcting types, normalizing/scaling, and aggregating.

What are common strategies for handling missing values before visualization?

Deletion (listwise/pairwise), imputation (mean, median, mode, or model-based), or explicitly encoding 'missing' as a category so gaps aren't silently hidden.

What is normalization vs. standardization in preprocessing?

Normalization rescales to a fixed range, often $[0,1]$ via $x' = \frac{x - x_{\min}}{x_{\max} - x_{\min}}$; standardization rescales to zero mean and unit variance via $z = \frac{x - \mu}{\sigma}$.

Why aggregate data before visualizing large datasets?

To reduce overplotting and noise, reveal higher-level patterns, and improve rendering performance by summarizing many points into meaningful groups (e.g., binning, averaging).

What is annotation in a visualization and why use it?

Adding explanatory text, callouts, reference lines, or markers directly on a chart to highlight key points, provide context, and guide the viewer's interpretation.

What are best practices for labeling chart axes?

Include clear titles, units of measurement, readable tick intervals, and direct data labels where helpful; avoid rotating text excessively and avoid clutter.

What is direct labeling and why is it often preferred over a legend?

Placing labels next to the data elements (e.g., at the end of lines) rather than in a separate legend, reducing the eye movement and cognitive load needed to match colors to categories.

What defines an interactive visualization?

A visualization that lets users manipulate it — through filtering, zooming, panning, hovering for tooltips, brushing, and linking — to explore data dynamically rather than viewing a static image.

What is brushing and linking in interactive visualization?

Selecting (brushing) data in one view automatically highlights or filters the corresponding data in other linked views, enabling multidimensional exploration.

What is the 'overview first, zoom and filter, then details-on-demand' principle?

Shneiderman's Visual Information-Seeking Mantra, describing the ideal workflow for interactive exploration of complex datasets.

What is geospatial visualization?

Representing data tied to geographic locations on maps, using elements like points, choropleths, and heat layers to reveal spatial patterns.

What is a choropleth map?

A map in which geographic regions are shaded or colored in proportion to a statistical variable (e.g., population density), typically using a sequential or diverging color scale.

Why should choropleth maps usually display normalized rates rather than raw counts?

Because larger or more populous regions naturally have higher raw totals; normalizing to rates or per-capita values prevents area/population size from confounding the pattern.

What challenges are unique to 3D visualization?

Occlusion (objects hiding others), perspective distortion of sizes, difficulty judging depth and exact values, and the need for rotation/interaction — so 3D is justified mainly for inherently spatial/volumetric data.

How is data visualization used in business analytics and healthcare?

Business analytics: KPI dashboards, sales/funnel and cohort analysis to drive decisions. Healthcare: patient monitoring dashboards, epidemiological maps, and visualizing trial or diagnostic data to spot trends and outliers.

How is data visualization applied in public policy?

Communicating census, economic, and public-health data to policymakers and citizens — e.g., dashboards and infographics for transparency, resource allocation, and evidence-based decision-making, with emphasis on honest, accessible design.