|
Activity Number:
|
329
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #308349 |
|
Title:
|
Principal Component Analysis for Data Reduction of Gait Analysis for Knee Prosthesis Research
|
|
Author(s):
|
Jeffrey Murphy*+ and Luke Aram and Jordan Lee and Travis Bennett and Catherine Truxillo and Paul Rullkoetter
|
|
Companies:
|
DePuy Orthopaedics and DePuy Orthopaedics and DePuy Orthopaedics and DePuy Orthopaedics and SAS Institute Inc. and University of Denver
|
|
Address:
|
700 Orthopaedic Drive, Warsaw, IN, 46582,
|
|
Keywords:
|
principal components analysis ; knee prosthesis design ; DOE ; T-square ; scree plot ; Gait analysis
|
|
Abstract:
|
Principal components analysis is useful in reducing voluminous gait data into manageable amounts while retaining much of the original information. Thirty factors in knee prosthesis design were set at two levels in a resolution IV fractional factorial DOE. Finite element analysis virtually built each run. Fifteen responses were measured. Principal components (PCs) were retained using the scree plot elbow method. Varimax orthogonal rotation and waveform plots were used when multiple PCs were retained. Influential runs were identified using T-square statistics. Between one and four PCs were retained for each response. Sagittal major radius and femoral internal/external rotation were the two most significant factors. Level setting combinations were carefully reviewed on influential runs. This was successful in reducing gait data into the critical few factors vital to knee prosthesis design.
|
