Exploratory Data Analysis

Also known as · EDA

A first-pass investigation of a dataset with summaries and charts, before any modelling, to find structure, problems and hypotheses. The lecture calls it a doctor's checkup: check the vitals before prescribing treatment.

When to use

Follow the reusable workflow: load and inspect, check types, summary statistics (mean versus median to flag skew), check quality (missing values, duplicates, outliers), visualise, then hypothesise. Most business questions are answered by EDA alone, and Anscombe's quartet shows summary statistics are never enough without plotting.

See DS Lec 2.

Appears in