Types of Data

Data is classified along several independent axes. Every axis matters, because the type of data decides which statistical measure and which chart is valid. Using a mean on nominal data or a pie chart on continuous data is a classic mistake.

1. Classification by Structure

# STRUCTURED — every record has the same fields
students = [
    {"roll": 101, "name": "Riya",  "marks": 88},
    {"roll": 102, "name": "Zoya",  "marks": 76},
]

# SEMI-STRUCTURED — JSON: self-describing, but records may differ
order = {
    "id": "ORD-9012",
    "items": [{"sku": "A1", "qty": 2}, {"sku": "B7", "qty": 1}],
    "gift_wrap": True          # this key may be absent in other orders
}

# UNSTRUCTURED — free text; no fields at all
review = "Delivery was quick but the packaging was torn on one side."

2. Classification by Nature

Qualitative (Categorical) data describes qualities or categories — it labels rather than measures.

Quantitative (Numerical) data describes quantities — it can be counted or measured, and arithmetic on it is meaningful.

Discrete vs Continuous:

3. Classification by Measurement Scale — NOIR

This is the most examined classification. Remember the mnemonic NOIR: Nominal, Ordinal, Interval, Ratio.

Why "true zero" matters: 40 °C is not "twice as hot" as 20 °C, because 0 °C is an arbitrary reference point, not an absence of temperature — so temperature in Celsius is interval. But ₹40,000 is exactly twice ₹20,000, because ₹0 means genuinely no money — so income is ratio.

import pandas as pd

df = pd.DataFrame({
    "student":   ["Riya", "Zoya", "Kabir", "Aarav"],   # Nominal
    "grade":     ["A", "B", "A", "C"],                 # Ordinal
    "temp_c":    [36.8, 37.2, 36.5, 38.1],             # Interval
    "marks":     [88, 76, 91, 54],                     # Ratio
})

print(df["student"].mode()[0])   # nominal -> mode only
print(df["grade"].value_counts()) # ordinal -> counts/median rank
print(df["marks"].mean())         # ratio -> mean is valid: 77.25
# df["student"].mean()  ->  MEANINGLESS: never average a nominal column

4. Classification by Source

5. Other Important Data Types in Analytics

Choosing Techniques by Data Type

Getting this classification right in the first lesson of a project saves an enormous amount of rework later — it determines your cleaning strategy (Unit 1), your summary statistics and charts (Unit 2), and even which algorithm can legally consume the column (Unit 3).