Parametric and Nonparametric Repeated-Measure Analysis of Growth Patterns

• Activity Number: 363
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section*
• Author(s): Xueying Li*+ and Sandra Rodriguez-Zas+ and Johnathan Beever and Michael Ellis and Floyd McKeith and Barbara Bailey
• Affiliation(s): University of Illinois, Urbana-Champaign and University of Illinois, Urbana-Champaign and University of Illinois, Urbana-Champaign and University of Illinois, Urbana-Champaign and University of Illinois, Urbana-Champaign and University of Illinois, Urbana-Champaign
• Keywords: spline ; nonlinear models ; mixed effects ; genetics

Abstract

A study of longitudinal weight measurements from 1100 genetically related pigs from birth to slaughter was conducted. The data consisted of multiple measurements correlated on time and genetic make-up repeated over multiple time points. Parametric approaches including linear (orthogonal and non-orthogonal polynomials on age), and nonlinear mixed effects models were evaluated. A spline smoothing nonparametric function was implemented using a linear mixed model approach. An Average Information REML approach provided estimates of variance components in the available unbalanced data set with missing observations. Hypothesis testing was based on F and likelihood ratio tests statistics. The impact of accounting for part or the complete genetic relationship on the estimates and inferences was explored. The analyses showed that the growth pattern was significantly influenced by fixed effects covariates including sex, parity, and litter size. More complex models enhanced the goodness of fit at the cost of higher dimensionality and, in some cases, estimability problems. Failure to account for all known genetic relationships lead to biased estimates.