7.3 Guiding Principles in the Search for Interactions

Statistical experimental design to establish casual relationships between independent and dependent variables, foresees:

• control
• randomization
• replication

Interactions can be of different degrees:

The identification of the interactions can be challenging, and even more challenging can be the identification of the shepherd interaction effects.

The framework for identifying significant interactions (Wu ans Hamada 2011) for experimental design and predictive modeling is based on:

• interaction hierarchy (degree of interaction)

• effect sparsity (only a fraction of the interaction effects can be effective)

• effect heredity (implies significant factors preceding interaction explain the most of the response)

  • strong heredity (interaction only with significant preceeding factors)
  • weak heredity (any interaction with one significant factor)

High order interaction happen in real life data (interactions among species).