Multivariate for Statisticians
Multivariate methods analyse several variables at once to find structure. Principal component analysis reduces dimensions while keeping variance, factor analysis models latent traits, and clustering groups similar observations. Discriminant analysis and MANOVA handle group comparisons across many outcomes. They depend on covariance, correlation and linear algebra.
Stage: Modelling · Part of the Statistician roadmap · Last reviewed 19 November 2026
What to learn in Multivariate
- Standardise variables and run principal component analysis
- Read a scree plot to decide how many components to keep
- Cluster observations with k-means and hierarchical clustering
- Interpret a covariance and correlation matrix for related variables
- Classify observations with linear discriminant analysis
Multivariate interview questions
Why standardise variables before PCA?
PCA finds directions of maximum variance, so a variable measured on a large scale, like income in rupees, would dominate one on a small scale. Standardising to mean zero and unit variance gives each variable equal weight, or you can use the correlation matrix instead.
How do you choose the number of clusters in k-means?
Try several values of k and compare with the elbow method on within-cluster variance, the silhouette score or the gap statistic. No single criterion is decisive, so also check whether the clusters are stable and meaningful for the problem.
Where Multivariate fits in the Statistician roadmap
Multivariate belongs to the Modelling stage (intermediate level). Relate variables.
Other units in this stage
Back to the full Statistician roadmap