Sorghum is the leading cereal crop worldwide and the top five cereal crops in production annually. Yields of sorghum are affected by different factors from which genotype and genotype by environmental interaction are the main sources of variability of yield. The study aims to evaluate genotype performance and examine the nonlinear dependence of grain yield and yield-related traits of sorghum using a generalized additive mixed model and also identify the impact of treatment on grain yield. The data was collected using the lattice square design of the experiment to characterize the yield and yield-related traits. The study considered grain yield as a response and incorporated yield-related traits as fixed, genotype, genotype by treatment and replication within the treatment as random effects and smooth terms for the analysis. The analysis was performed through GAMM. The result shows that there was a significant difference in grain yield among the treatments and that most of the variability of the grain yield (more than 84%) was explained by genotype, genotype by treatment interaction whereas the variability of grain yield (less than 2%) due to smooth term and replication within treatment was negligible. The study found that grain yield and plant height had significant non-linear dependence and also grain yield and panicle width was non-linearly associated where grain yield and other continuous yield-related traits were linearly associated. The relationship between grain yield and the independent variables was non-linear Genotypes that are recommended for mass production in a cultivate are G196, G149 and G187 respectively in order of performance, Genotypes which enabled to perform better under irrigated condition are G145, G149 and G142 whereas genotype that have better performance under stress condition are G196, G22 and G19 in order of performance. In general, the study recommends using GAMM for the genotype selection of crops and identifying linear or non-linear associations between grain yield and continuous covariates of yield related traits in crop production. The future work will focus on multivariate analysis to perform genotype selection for crop production.
