1.3 Terminology

• Unsupervised models are used to understand relationships between variables or sets of variables without an explicit relationship between variables and an outcome.
  • Examples: PCA, clustering, autoencoders.
• Supervised models have an outcome variable.
  • Examples: linear regression, decision trees, neural networks.
  • Regression: numeric outcome
  • Classification: ordered or unordered qualitative values.
• Quantitative data: numbers.
• Qualitative (nominal) data: non-numbers.
  • Qualitative data still might be coded as numbers, e.g. one-hot encoding or dummy variable encoding
• Data can have different roles in analyses:
  • Outcomes (labels, endpoints, dependent variables): the value being predicted in supervised models.
  • Predictors (independent variables): the variables used to predict the outcome.
  • Identifiers

Choosing a model type will depend on the type of question we want to answer / problem to solve and on the available data, among other things.