Univariate Analysis — Continuous and Categorical — Data Analysis & Feature Engineering

Univariate analysis is the examination of each variable in isolation, the stage of EDA where you build a complete picture of every column's behaviour before considering how columns relate. It exists because you cannot understand relationships between variables until you understand each variable on its own — its centre, spread, shape, range, and frequency of categories. Without thorough univariate analysis, analysts miss the skew that will distort a model, the rare category that will break an encoder, or the impossible value that signals a data error. Univariate analysis differs fundamentally for continuous variables, where distributions and outliers matter, versus categorical variables, where frequencies and cardinality matter, and knowing how to interrogate each type is foundational to all downstream feature engineering and modelling.

Analogy🏏Cricket

🏏 Think of it like cricket: Reducing twenty batting statistics to two dimensions with PCA produces a flat projection that may look like one undifferentiated cloud, while t-SNE or UMAP finds that the data actually organises into distinct clusters — aggressive pinch-hitters, steady anchors, explosive finishers — whose separation PCA's flat projection smeared together. Just as the non-linear reduction reveals the natural groupings that flat projection could not, t-SNE and UMAP reveal structure that PCA misses because it can only flatten, not curve to follow the data's natural shape. The insight is that non-linear reduction methods follow the data's true curved geometry rather than forcing a flat projection, revealing the cluster structure and local neighbourhoods that linear methods like PCA cannot preserve.