An Information Theoretical Method for Analyzing Unreplicated Designs with Binary Response

Authors

  • Krystallenia Drosou National Technical University of Athens
  • Christos Koukouvinos National Technical University of Athens

DOI:

https://doi.org/10.57805/revstat.v17i3.273

Keywords:

two-level factorial designs, unreplicated experiments, generalized linear models, symmetrical uncertainty

Abstract

The analysis of unreplicated factorial designs constitutes a challenging but difficult issue since there are no degrees of freedom so as to estimate the error variance. In the present paper we propose a method for screening active effects in such designs, assuming Bernoulli distributed data rather than linear; something that hasn’t received much attention yet. Specifically, we develop an innovating algorithm based on an information theoretical measure, the well-known symmetrical uncertainty, so that it can measure the relation between the response variable and each factor separately. The powerfulness of the proposed method is revealed via both, a thorough simulation study and a real data set analysis.

Published

2019-07-09

How to Cite

Drosou , K., & Koukouvinos , C. (2019). An Information Theoretical Method for Analyzing Unreplicated Designs with Binary Response. REVSTAT-Statistical Journal, 17(3), 383–399. https://doi.org/10.57805/revstat.v17i3.273

Most read articles by the same author(s)