Title: Correlation modelling of complex data – physics, statistics and heuristics

Authors: Kenneth W. Jackson, Mahmood Tabaddor

Addresses: Department of Industrial Engineering Technology, Southern Polytechnic State University, Marietta, Georgia 30060, USA. ' Corporate Research, Underwriters Laboratories Inc., Northbrook, Illinois 60062, USA

Abstract: In this paper, we cover some principles and guidelines that are useful for modelling and interpreting data associated with highly complex physical phenomena such as occur in multidisciplinary fields. We compare and contrast the theoretical and statistical-empirical modelling paradigms and discuss how they interact and are complementary. Using an example taken from the field of fire engineering, we review how the approach can influence the efficiency and effectiveness of experimental or numerical investigations. We show how integrating dimensional analysis with experimental design techniques and regression modelling can reduce experimentation schedules and costs and improve insight. We further illustrate several useful strategies and caveats for modelling highly complex data. We describe some common limitations and misconceptions of data analysis along with features of graphical representation that can facilitate interpretation.

Keywords: dimensional analysis; complex data; fire engineering; correlation modelling; complex physical phenomena; physics; experimental design; regression modelling.

DOI: 10.1504/IJDATS.2010.037476

International Journal of Data Analysis Techniques and Strategies, 2010 Vol.2 No.4, pp.336 - 355

Published online: 14 Dec 2010 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article