Measurement Error Model for Shape Data

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Jiejun Du*+
• Companies: University of South Carolina
• Keywords: Measurement Error ; Shape Data ; Structural Model

Abstract

In statistical shape analysis the calculation of a non- Euclidean distance between two shapes or the matching of objects using Procrustes analysis can be formulated by using a complex linear regression equation for two dimensional landmark data. However, conventionally there is only one error term in the regression equation and measurement error is ignored. In reality, shape data are often prone to measurement error and so in this study we consider measurement error models for shape data. First we show that estimation of some parameters in the naive ordinary least squares estimator is biased when the measurement error is ignored. Then we consider structural measurement error models for shape data and derive improved estimators for the parameters in the model. The methodology is illustrated with an application in forensic face identification.