Follow the transformation
Inspect schema, unit of analysis and target definition.
Measure missingness, duplicates, class balance and impossible values.
Study univariate distributions and subgroup differences.
EDA & Data Quality · Lesson 21
Univariate Analysis
Univariate Analysis is a data-understanding step that examines whether the dataset accurately represents the problem you intend to model.
What Univariate Analysis actually means
Univariate Analysis is a data-understanding step that examines whether the dataset accurately represents the problem you intend to model. EDA is not merely plotting; it is a search for structure, quality problems and leakage risks that can invalidate later evaluation.
Univariate Analysis matters because model quality cannot exceed the meaning and integrity of its data. Profiling, EDA, label checks and leakage checks expose problems that a sophisticated algorithm may otherwise exploit or hide.
Deeper walkthrough
Read Univariate Analysis as a mechanism, not a recipe
Treat this as a sequence of observable decisions rather than one opaque command. Stage 1: Inspect schema, unit of analysis and target definition. Stage 2: Measure missingness, duplicates, class balance and impossible values. Stage 3: Study univariate distributions and subgroup differences. Final checkpoint: Audit timestamps and data provenance for information that would not exist at prediction time.
Know what would convince you
- Compare missingness, distributions, row counts and key constraints before and after the change.
- Inspect representative changed rows and confirm the rule with domain/data documentation.
Useful distinctionQuestion: Ask what the operation is intended to answer.
Click a stage to inspect what happens, what changes, and what should be checked before moving on.
Stage 1
Inspect schema
Inspect schema, unit of analysis and target definition. For Univariate Analysis, make this checkpoint explicit by recording the evidence inspected, the expected result, and the condition that would make you reject the current result.
Verification focus: record the evidence you inspected and the condition that would make this stage fail.
Trace the mechanism step by step
- Inspect schema, unit of analysis and target definition.
- Measure missingness, duplicates, class balance and impossible values.
- Study univariate distributions and subgroup differences.
- Use bivariate/multivariate views to identify associations, nonlinearities and confounding structure.
- Audit timestamps and data provenance for information that would not exist at prediction time.
Make the concept concrete
Demonstration
Python example
# Step 1 — Import the module so its functions/classes are available to the rest of this example.
import numpy as np
# Step 2 — Construct `x` as an array so vectorised numerical operations can be applied consistently.
x = np.array([1,2,2,3,3,3,4,9])
# Step 3 — Display the current value explicitly so the result/state can be inspected during execution.
print("min/max:", x.min(), x.max())
# Step 4 — Display the current value explicitly so the result/state can be inspected during execution.
print("median:", np.median(x))
# Step 5 — Display the current value explicitly so the result/state can be inspected during execution.
print("q1/q3:", np.quantile(x,[.25,.75]))Expected / illustrative result
min/max: 1 9; median: 3; quartiles show the central mass while 9 creates a long right tail.
Interpret the result.
For Univariate Analysis, trace representative source rows/columns into the result and reconcile row counts, dtypes, keys or missing values that the operation could change.
Know what this is — and what it is not
QuestionAsk what the operation is intended to answer.MechanismTrace the rule from input to output.EvidenceInspect a value, table, plot, error or metric that can falsify your expectation.When it is appropriate
Use Univariate Analysis when it helps diagnose, document or correct a data-quality issue without destroying information needed for the downstream question.
When to stop or reconsider
Do not “clean” automatically when the apparent anomaly may carry signal, reflect data collection, or require domain adjudication; preserve an audit trail of changes.
Failure modes to recognise
- Treating missing values, outliers or labels as purely technical defects without investigating how they were generated.
- Applying learned cleaning/imputation using information from validation/test data.
- Changing values without recording which rows changed and how distributions/counts were affected.
How to check the result
- Compare missingness, distributions, row counts and key constraints before and after the change.
- Inspect representative changed rows and confirm the rule with domain/data documentation.
- In predictive work, fit learned cleaning only on training data and verify the pipeline reproduces that boundary.
Demonstrate understanding
Try this:
Build a tiny, inspectable example of Univariate Analysis. First inspect schema, unit of analysis and target definition. Then measure missingness, duplicates, class balance and impossible values. Write the expected result before running it, and explain one condition that would make the result misleading or invalid.
Start by measuring the problem, not fixing it. Keep a before/after table of counts or distributions and inspect the exact records affected by the rule.
Check reasoning, not memorisation
Before trusting a result from Univariate Analysis, which check provides the strongest evidence that you understand and applied it correctly?
Keep the important distinctions visible
Step 1Inspect schema, unit of analysis and target definition.Step 2Measure missingness, duplicates, class balance and impossible values.Step 3Study univariate distributions and subgroup differences.Step 4Use bivariate/multivariate views to identify associations, nonlinearities and confounding structure.What to remember
- Univariate Analysis is a data-understanding step that examines whether the dataset accurately represents the problem you intend to model. EDA is not merely plotting; it is a search for structure, quality problems and leakage risks that can invalidate later evaluation.
- Inspect schema, unit of analysis and target definition.
- Treating missing values, outliers or labels as purely technical defects without investigating how they were generated.
- Compare missingness, distributions, row counts and key constraints before and after the change.