Notes
🔍

Qualitative Data

Some data tells you “how much,” like a house price or a blood pressure reading. Other data tells you “what kind,” like the neighborhood name or whether a tumor is benign or malignant—that second kind is qualitative data.

What it is
Qualitative data (often called categorical data) records qualities, labels, or categories rather than numeric measurements. The values are typically words or codes, but the key point is that the categories represent distinct groups, not amounts on a meaningful numeric scale.

Main subtypes
There are two common flavors:

  • Nominal: categories with no natural order (e.g., “red/green/blue,” “cat/dog,” “country”).
  • Ordinal: categories with a meaningful order, but unclear or uneven spacing (e.g., “low/medium/high,” 1–5 star ratings, disease stage I–IV).

Everyday examples
In real datasets, qualitative variables show up constantly:

  • House sales: neighborhood, property type (apartment/house), heating type.
  • Education: letter grade (A–F), school district, course subject.
  • Healthcare: diagnosis code, smoker/non-smoker, symptom present/absent.

Why it matters in AI/ML
Most ML models expect numbers, so qualitative data usually needs preprocessing. Common approaches include:

  • One-hot encoding for nominal categories (creates 0/1 indicator features).
  • Ordinal encoding for ordered categories (careful: it implies an order).
  • Target encoding or embeddings for high-cardinality features (like thousands of ZIP codes).

If you treat qualitative categories as ordinary numbers (e.g., “red=1, green=2, blue=3”), many models will incorrectly assume distances and averages that don’t exist, which can harm performance and fairness.

Qualitative Data are non-numeric observations that describe categories or attributes, such as labels, names, or types. They can be nominal (no inherent order) or ordinal (ordered categories). In statistics and ML, qualitative data often appear as categorical features or class labels and require encoding (e.g., one-hot or ordinal encoding) before many models can use them. Example: “color = red/green/blue” as an input feature for classification.

Imagine sorting your closet: you might group clothes by “shirts,” “pants,” or “jackets.” Those groups are names, not numbers. That’s the idea behind Qualitative Data: information that describes qualities or categories, like color (red/blue), brand (Nike/Adidas), or type (spam/not spam).

In statistics and AI, qualitative data helps a model tell things apart based on labels rather than measurements. A movie genre (“comedy,” “drama”) or a customer’s satisfaction (“happy,” “neutral,” “upset”) are common examples. Even though it isn’t numeric, it can still be very useful for making predictions and decisions.