|
Activity Number:
|
471
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #309671 |
|
Title:
|
The Entropy Correlation Coefficient and Its Application to Clinical Data Analysis
|
|
Author(s):
|
Ronald Pearson*+ and Jaakko Astola
|
|
Companies:
|
ProSanos Corporation and Tampere University of Technology
|
|
Address:
|
225 Market St, Harrisburg, PA, 17101,
|
|
Keywords:
|
association measures ; entropy correlation ; clinical data analysis
|
|
Abstract:
|
The entropy correlation coefficient (ECC) is a measure of association between nominal variables proposed by Astola and Virtanen in 1981. Recently, Craddock et al. suggested that the ECC is useful in the exploratory analysis of clinical data and that its properties deserve further study. This paper compares the performance of the ECC with four better-known alternatives, first for a set of simulation examples and then for three clinical datasets, including the one considered by Craddock et al. The simulation examples permit quantitative control of the degree of dependence between the two sequences, independent of potentially important factors like the number of levels for each nominal variable and their marginal distributions. The three clinical examples illustrate the range of practical problems where these measures are potentially useful.
|
