Understanding Data and Visualization Principles

DATA TYPE CLASSIFICATION:

Quantitative (Numeric):
  Continuous: Can take any value — Height, Revenue, Temperature
  Discrete:   Whole numbers only — Employees, Clicks, Orders

Categorical (Non-numeric):
  Nominal:  No order — City names, Colors, Product types
  Ordinal:  Has meaningful order — Ratings (1-5), Education level

Temporal (Time-based):
  Dates, timestamps, durations

Geographic:
  Countries, coordinates, regions

+─────────────────+────────────────────────────────────────────+
│ Data Type       │ Best Chart                                 │
+─────────────────+────────────────────────────────────────────+
│ Continuous qty  │ Histogram, KDE, Box plot, Line chart       │
│ Discrete qty    │ Bar chart, Dot plot                        │
│ Nominal cat     │ Bar chart, Treemap, Word cloud             │
│ Ordinal cat     │ Bar chart (ordered), Diverging bar         │
│ Two quantities  │ Scatter plot, Bubble chart, Heatmap        │
│ Time series     │ Line chart, Area chart, Candlestick        │
│ Part-of-whole   │ Pie (≤5), Donut, Treemap, Waffle           │
│ Geographic      │ Choropleth, Dot map, Bubble map            │
+─────────────────+────────────────────────────────────────────+

Pre-attentive attributes are processed by the brain in < 250ms
(before conscious attention):

✅ Most Powerful Encodings:
  1. Position      → Most accurate for quantitative comparison
  2. Length        → Second most accurate (bar charts)
  3. Color hue     → Best for categorical distinction
  4. Color value   → Best for showing magnitude (dark = more)

⚠️ Less Accurate Encodings (use with caution):
  5. Size/Area     → Hard to compare precisely (bubble charts)
  6. Angle         → Why pie charts are hard to read accurately
  7. Shape         → Good for categories, bad for quantity
  8. Texture       → Last resort only

Rule: Use POSITION and LENGTH for quantitative data.
      Use COLOR HUE for categorical distinction.

What do you want to show?
│
├── HOW THINGS CHANGE OVER TIME?
│   └── Line chart (continuous) or Bar chart (discrete periods)
│
├── HOW THINGS COMPARE?
│   ├── Few items (≤7) → Bar chart (horizontal for long labels)
│   └── Many items    → Sorted bar, Dot plot
│
├── HOW THINGS DISTRIBUTE?
│   ├── Single variable → Histogram or Box plot
│   └── By category    → Violin or grouped Box plot
│
├── HOW THINGS RELATE?
│   ├── Two variables  → Scatter plot
│   └── Many variables → Correlation heatmap or Pair plot
│
├── PART-OF-WHOLE?
│   ├── ≤5 categories  → Pie or Donut chart
│   └── Many categories→ Treemap or Stacked bar
│
└── WHERE THINGS ARE?
    └── Map (Choropleth for values, Dot for locations)

import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import numpy as np

fig, axes = plt.subplots(1, 3, figsize=(15, 4))

# Sequential palette — for ordered/magnitude data
cmap_seq = plt.cm.Blues
vals = np.linspace(0, 1, 8)
for i, v in enumerate(vals):
    axes[0].bar(i, 1, color=cmap_seq(v), edgecolor='none')
axes[0].set_title('Sequential (Blues)\nUse for: quantities, intensity')
axes[0].set_xticks([])

# Diverging palette — for data with meaningful midpoint
cmap_div = plt.cm.RdBu
for i, v in enumerate(np.linspace(0, 1, 8)):
    axes[1].bar(i, 1, color=cmap_div(v), edgecolor='none')
axes[1].set_title('Diverging (RdBu)\nUse for: correlation, profit/loss')
axes[1].set_xticks([])

# Qualitative palette — for categorical data
colors = ['#E41A1C','#377EB8','#4DAF4A','#984EA3','#FF7F00','#A65628']
for i, c in enumerate(colors):
    axes[2].bar(i, 1, color=c, edgecolor='none')
axes[2].set_title('Qualitative\nUse for: distinct categories')
axes[2].set_xticks([])

plt.tight_layout()
plt.savefig('color_palettes.png', dpi=150)
plt.show()

Gestalt Principles (how we group visual elements):

1. Proximity    → Elements close together appear related
                  → Group related chart elements (axes, legends)

2. Similarity   → Similar-looking elements seem part of same group
                  → Use same color for same category across charts

3. Continuity   → Eyes follow paths and lines
                  → Line charts guide eyes through time

4. Closure      → Brain fills in incomplete shapes
                  → Area charts feel "complete" even if truncated

5. Figure/Ground → We separate objects from background
                  → Use white space and gridlines carefully

Application:
  Keep legends close to the data they describe.
  Use consistent colors for same categories across all charts in a report.
  Add white space — crowded charts violate closure principle.