Data visualization Study Guide

📖 Core Concepts Data Visualization – Graphic representation (static, dynamic, interactive) of quantitative & qualitative data to help audiences explore, understand, and gain insights. Information Visualization – Focuses on large, complex datasets; adds value, improves cognition, and aids decision‑making via graphical displays. Narrative Visualization – Combines charts, maps, etc., with a story structure to guide the viewer through trends and relationships. Design Goal – Show accurate, up‑to‑date data simply; use aesthetically meaningful shapes and colors; avoid distraction. Data‑Ink Ratio – Maximize the proportion of ink that represents data; eliminate chartjunk (unnecessary 3‑D effects, extra legends, decorative elements). Pre‑Attentive Attributes – Visual features instantly perceived (length, orientation, color hue, shape). Use them for rapid comparisons. Bijective Mapping – One‑to‑one correspondence between visual element and data variable; improves readability. Variable Types – Categorical: nominal (no order) or ordinal (ordered). Quantitative: continuous (infinitely precise) or discrete (finite set). Visualization Literacy – Ability to read, critique, and create effective visual displays; essential for accurate interpretation. Visual Analytics – Integration of statistical analysis, interactive visual interfaces, and human reasoning to draw conclusions. --- 📌 Must Remember Tufte’s Core Principles – No distortion, encourage data‑centric thinking, show many numbers in small space, enable eye comparison, support multi‑level detail, serve a clear purpose. Data‑Ink Maximization – Remove all non‑data ink; every pixel should convey information. Chart Selection by Message | Message | Best Chart | |---------|------------| | Time‑Series change | Line chart | | Ranking / order | Bar chart | | Part‑to‑Whole | Pie / Stacked bar (few categories, large differences) | | Deviation from reference | Bar chart (actual vs. target) | | Frequency distribution | Histogram or Box plot | | Correlation | Scatter plot | | Nominal comparison | Bar chart | | Geographic comparison | Choropleth / Cartogram | Pre‑Attentive Preference Order – Length > angle > position > area > volume > color saturation. Avoid: Using area (pie slices) to compare precise values; starting Y‑axis above zero without justification; excessive colors for categorical data; 3‑D effects that hide true values. Log Scale – Use when data span several orders of magnitude (exponential growth/decay). --- 🔄 Key Processes Define Purpose & Audience What decision or insight is needed? Who will view it (technical vs. non‑technical)? Identify the Quantitative Message (time‑series, ranking, etc.). Choose the Optimal Chart Type (use the table above). Map Variables to Encodings Position → quantitative (most accurate) Length → ordinal/quantitative (bars) Color hue → categorical Size → quantitative (but less precise) Maximize Data‑Ink Ratio Strip gridlines, background patterns, unnecessary legends. Add Supporting Text Clear title, axis labels, concise caption highlighting the key takeaway. Iterative Testing Show to a sample of the target audience → collect feedback → refine. Add Interactivity (if needed) Brushing → highlight linked points across plots. Linking → synchronize selections. Scaling/Zoom → explore dense regions. --- 🔍 Key Comparisons Data Viz vs. Information Viz vs. Narrative Viz Data Viz: raw data → visual; focus on accuracy. Info Viz: large/complex data → insight; emphasizes cognition. Narrative Viz: data + storytelling → persuasive flow. Statistical Graphics vs. Data Visualizations Statistical: exploratory, researcher‑centric, less emphasis on visual appeal. Data: audience‑centric, aesthetic, storytelling. Bar Chart vs. Pie Chart Bar: encodes length → high precision. Pie: encodes angle/area → low precision; only for few, markedly different slices. Static vs. Interactive Visuals Static: good for printed reports, simple messages. Interactive: enables brushing, linking, zoom – essential for large, multi‑dimensional data. --- ⚠️ Common Misunderstandings “More colors = clearer” – Excessive hues overload pre‑attentive processing; stick to a limited palette. Pie charts are always appropriate for percentages – They hide precise comparisons; use bar or stacked bar instead. Chartjunk improves engagement – Decorative elements can obscure data and reduce data‑ink ratio. Log scales are always safe – They can mislead if axis breaks aren’t clearly indicated. Higher resolution always means better – Over‑resolution adds noise; focus on clarity, not pixel count. --- 🧠 Mental Models / Intuition “Length beats area” – Imagine comparing two sticks vs. two circles; sticks (bars) give an immediate sense of magnitude. “Pre‑attentive Fast Path” – The brain scans for length, orientation, or color hue first; design to let the answer pop out without deliberate thought. “Data‑Ink as Signal‑to‑Noise” – Treat every non‑data pixel as noise; the clearer the signal, the faster the insight. --- 🚩 Exceptions & Edge Cases Pie Chart Use – Only when ≤ 5 categories, one dominates, and the audience is non‑technical. Log Scale – Must label tick marks with actual values (e.g., 10, 100, 1 000) and note “log scale” in the caption. Color Hue for Quantitative Data – Avoid; use sequential palettes (light→dark) instead of categorical hues. Small Multiples – When a single chart becomes cluttered, split into a grid of simpler charts. 3‑D Bar/Column Charts – Acceptable only for decorative presentations; never for precise comparison. --- 📍 When to Use Which | Situation | Recommended Visual | |-----------|---------------------| | Show trend over months/years | Line chart (add markers for key events) | | Compare sales across product categories | Bar chart (horizontal if many categories) | | Display market share of 3‑5 firms | Pie chart (large differences) | | Highlight outliers & distribution | Box plot (with jittered points) | | Explore relationships between two continuous variables | Scatter plot (add trend line if needed) | | Map disease incidence by county | Choropleth map (consider cartogram for population‑adjusted view) | | Show project schedule & dependencies | Gantt chart | | Visualize hierarchical file system | Treemap or Dendrogram | | Need interactive drill‑down on dense scatter | Linked brushing & scaling | | Communicate a story with sequential steps | Narrative visualization (combine charts, captions, and flow arrows) | --- 👀 Patterns to Recognize Straight, upward line → consistent growth; sharp bends → regime change. Long tail in histogram → skewed distribution; multiple peaks → multimodal. Parallel lines in parallel‑coordinates → strong correlation across dimensions. Large, isolated points in scatter → outliers; clusters of same color/shape → groups. Heat map gradient – smooth gradient = gradual change; abrupt color change = boundary. Sankey width variation – larger flow = higher volume. --- 🗂️ Exam Traps Distractor: “Use a pie chart for ranking” – pie charts hide precise ordering; bar chart is correct. Distractor: “Start Y‑axis at 50 % to emphasize growth” – unless the baseline is contextually justified, this misleads perception. Distractor: “More colors make a map clearer” – too many hues create visual clutter; use a limited sequential or diverging palette. Distractor: “Log scale always shows trends better” – can mask absolute differences; ensure axis labeling is explicit. Distractor: “3‑D effect adds professionalism” – adds chartjunk and distorts length perception. Near‑miss answer: Choosing a stacked bar for part‑to‑whole when categories have many small slices; a treemap or donut chart may preserve readability. ---