Dealing with Correlated Observations in Time Using Repeated Measurements.

In many agricultural and biological situations, measurements and observations (e.g., data) are routinely collected on the same experimental unit over time. This is termed repeated measurements. In a repeated measurements study, there are numerous methods that could be applied to analyze such data, however, many of the methods are not correct. These incorrect methods include (i) treating time as a factor, (ii) analyzing each individual time point separately, or (iii) treating the experimental structure as a split plot. With an improper analysis, the results and inferences about the meaning of the study will most likely be biased. Therefore, it is critical that a proper analysis is considered for repeated measurements. In this talk, I will highlight how we can analyze repeated measurements using a mixed model framework. Using a case study, discussion will focus on how to construct a mixed model and select an appropriate correlation structure. Emphasis will be placed on the use of PROC MIXED in SAS, however, a brief discussion about PROC GLIMMIX and NLMIXED will be considered.