Analysis of a Long‐term Experiment with Repeated‐measurement Models

Abstract

Abstract Long‐term experiments are a classical case of repeated measurements. Traits are measured on the same experimental unit over many years so that correlations arise between the observations made on the same plot in consecutive years. This paper describes the analysis of a three‐crop‐rotation long‐term experiment. We analysed the yields of the crops and the organic carbon content in the topsoil over 30 consecutive years. Several variance–covariance approaches are discussed and the trait‐specific best fit is interpreted. Mixed models are used to describe the structure of the experiment. Both yields and soil organic carbon show a more or less pronounced variance heterogeneity. Especially for yields, the heterogeneity of cycles and years is dominant. The consideration of correlations results in a better model fit in all cases.