Mixed Model procedure is used in the modeling and estimation of treatment efforts and variance components. It extends the generalized linear model (ordinary ANOVA) by avoiding the restrictive assumption of the error terms in ANOVA through the incorporation of the variance-covariance structure of all random terms. The mixed model procedure uses generalized least squares rather than the ordinary least squares, and estimates from mixed procedure are always better than those from the generalized linear model. The implementation of mixed model in SAS provides a powerful and flexible tool for analysing a wide range of data, including - the analysis of repeated measures, the analysis of spatially correlated data, the analysis of non-orthogonal designed experiments, and analysis of multilevel or hierarchical data (e.g. as in. on-farm trials). This paper looks at the two most important designs used in on-farm trials, 1) nested (hierarchical) design, and 2) augmented block design, and presents SAS programs useful for analysing data resulting from these designs.

Smallholder Farmers; Plant Breeding; Plant Production; Farm Management; Diseases Control; Agribusiness; Genetic Improvement; Plant Genetic Resources; Livelihoods; Handling, Transport, Storage And Protection Of Agricultural Products

Augmented Design; Mixed Model; Nested Design; On-Farm Trials; Variance Components; Sas Program

Africa; West Africa

Benin; Nigeria