Many variables at once.
Answer five questions about Multivariate Analysis and get instant feedback.
Hidden factor that helps explain observed measurements
In psychology, traits like "intelligence" are treated as latent variables: you cannot measure them directly, but they reveal themselves through patterns across many test questions.
What predicts which category a new case belongs to?
Many classifiers output class probabilities, not just labels, so you can tune a decision threshold to trade off false alarms versus missed cases depending on what matters most.
What can change a lot when the chosen distance or similarity measure changes?
Switching from Euclidean to cosine distance can flip clusters from grouping by magnitude to grouping by direction, so the same data can yield totally different "natural" groupings.
This image question appears in the interactive quiz.
A confusion matrix separates correct classifications from specific types of mistakes.
What matrix equals a covariance matrix after each variable is standardized to mean $0$ and variance $?
Because the correlation matrix is scale-free, PCA on standardized data is equivalent to eigen-decomposing it, and its off-diagonals are exactly the cosines of angles between centered variable vectors.
Learn and play with MAGMA Mentor to build your understanding step by step.