|
Abstract:
An algorithm was developed using univariate statistics to reduce and analyze multivariate and multiple group data sets. The algorithm features the quantitative and selectivity figures of merit of receiver operating characteristics (ROC) curve methodology. This merging of two separate statistical analysis techniques resulted in the ability to address more than one variable in more than two experimental groups in a systematic fashion. The classic Fisher iris flower data set is treated as one variable and two cases at a time following conventional ROC curve methodology. Redundant, noisy, and low information containing variables are removed. The remaining information-rich variables are systematically merged using ROC curve techniques. The new algorithm using ROC curve techniques produces a master vector of down selected variables. The ROC curve technique can be used to process any data distribution whether linear or nonlinear; the inherent trend and fundamental nature of the raw data arc not compromised. No data set normalization or scaling procedures are necessary. Combining qualitative and quantitative aspects of data analysis into a univariate statistical method provides advantages in terms of algorithm understanding for the layman as well as enhanced computer efficiency and information-rich analysis.
| Limitations: |
 APPROVED FOR PUBLIC RELEASE |
| Description: |
Final rept. Dec 2009-Aug 2010 |
| Pages: |
45 |
| Report Date: |
Jan 2012 |
| Report Number: |
A984455 |
|
|
|
|
|
Keywords relating to this report:
Email This Abstract
