文档介绍：Statistical Analysis of Repeated Measures Data Using SAS Procedures1,2
R. C. Littell*,3, P. R. Henry†, and C. B. Ammerman†
Departments of *Statistics and †Animal Science, University of Florida, Gainesville 32611-0339
ABSTRACT: Mixed linear models were developed the SAS System include PROC MIXED. This proce-
by animal breeders to evaluate ic potential of dure implements random effects in the statistical
bulls. Application of mixed models has recently spread model and permits modeling the covariance structure
to all areas of research, spurred by availability of of the data. Thereby, PROC MIXED pute
puter software. Previously, mixed model efficient estimates of fixed effects and valid standard
analyses were implemented by adapting fixed-effect errors of the estimates. Modeling the covariance
methods to models with random effects. This imposed structure is especially important for analysis of
limitations on applicability because the covariance repeated measures data because measurements taken
structure was not modeled. This is the case with close in time are potentially more highly correlated
PROC GLM in the SAS System. Recent versions of than those taken far apart in time.
Key Words: Statistical Analysis, Direct parisons, Analysis of Variance
1998 American Society of Animal Science. All rights reserved. J. Anim. Sci. 1998. 76:1216–1231
Introduction ment. The feature of repeated measures experiments
that requires special attention in data analysis is the
The term “repeated measures” as used in this paper correlation pattern among the responses on the same
refers to multiple responses taken in sequence on the animal over time.
same experimental unit, such as an animal. Usually,
the responses are taken over time, as in weekly weight
measurements to establish growth curves. However, Methods for Analyzing Repeated
the repeated measures could be taken in spatial Measures Data
sequence, such as dimensions of vertebrae. The typical
repeated meas