Ridit Analysis for Cooper-Harper and Other Ordinal Ratings for Sparse Data - A Distance-based Approach
Abstract
Ordinal categorical data (OCD), such as opinion rankings, are common in many areas of application. In the Air Force, Cooper-Harper ratings are used extensively for the assessment of Flying Qualities. OCD is not, however, a ratio-scale measurement and cannot be treated as ordinary numbers. Notwithstanding this, the ordinal scores are often regarded as ratio-scale and analyzed incorrectly using means and variances. A method of correct analysis of OCD leading to statistically valid hypothesis tests and based on a method of probability scoring or Ridits, has found wide applicability for other large-data-set applications such as Epidemiology. This paper explains the use of Ridits and examines how we might effect a Ridit analysis on the often sparse data sets in many Flying Qualities applications. All flying qualities data in this paper is synthetic, and has been simulated to illustrate Ridit analysis. The method of this paper is to fit empirical Beta distributions to observed data, and then to use a randomization approach to make inferences on the difference between distributions based on a distance metric.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2016
- Accession Number
- AD1017266
Entities
People
- Arnon Hurwitz
Organizations
- Air Force Test Center