Skip to content

Dichotomous Key for Multivariate Statistics (Basic Version)

Follow each step in order. Each step either leads to a method or to the next step.

1. Is there a defined dependent (outcome) variable?

  • Yes → go to Step 2
  • No / exploratory analysis → go to Step 5

2. Type of dependent variable

3. Categorical dependent variable

  • Two categories → Logistic Regression
  • More than two categories → Multinomial Logistic Regression

4. Continuous dependent variable

  • One predictor → Simple Linear Regression
  • Multiple predictors (continuous and/or categorical) → Multiple Linear Regression (MLR)
  • Categorical predictors only → ANOVA

5. Exploratory analysis (no dependent variable)

  • Goal: Reduce dimensionality → go to Step 6
  • Goal: Group observations (clustering) → go to Step 7
  • Goal: Classify observations into known groups → LDA (Linear Discriminant Analysis)

6. Dimension reduction

  • Variables continuous, Euclidean distance appropriate → PCA (Principal Component Analysis)
  • Distance/dissimilarity matrix (e.g., Bray-Curtis, UniFrac) → PCoA (Principal Coordinates Analysis) or NMDS

7. Clustering

  • Number of clusters known → k-Means Clustering
  • Number of clusters unknown, explore structure → Hierarchical Clustering

Basic Abbreviations

  • PCA = Principal Component Analysis — linear ordination with Euclidean distances
  • PCoA = Principal Coordinates Analysis — ordination using any distance matrix
  • NMDS = Non-metric Multidimensional Scaling — rank-based ordination
  • LDA = Linear Discriminant Analysis — classification into known groups
  • MLR = Multiple Linear Regression
  • ANOVA = Analysis of Variance

Notes

  • This simplified key covers the most commonly used multivariate methods
  • Always check assumptions before applying methods
  • For more advanced methods (constrained ordination, MANOVA, PERMANOVA, etc.), see the extended version