Extended Case Study of Causal Learning within Architecture Research (preliminary results)

Abstract

Regression must be interpreted in a context of a DAG! Correlation, hence regression, may be fooled by spurious association!Before jumping into regression, we need a Directed Acyclic Graph (DAG) representing our context.We then need to determine which paths are causal and which are spurious.We then must block spurious correlation paths.Lastly, we then conduct regression with the correct set of factors!

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 31, 2018
Accession Number
AD1084133

Entities

People

  • David Danks
  • Michael D. Konrad
  • Rick Kazman
  • Robert W. Stoddard

Organizations

  • Carnegie Mellon University

Tags

DTIC Thesaurus Topics

  • Algorithms
  • Case Studies
  • Causal Reasoning
  • Engineering
  • Equations
  • Greenhouses
  • Learning
  • Schools
  • Software Design
  • Software Development
  • Universities

Fields of Study

  • Computer science

Readers

  • Graph Algorithms and Convex Optimization.
  • Neural Network Machine Learning.
  • Regression Analysis.