Model comparisons for genetic evaluation of gait type in Mangalarga Marchador horses

• Autor(a) principal: Bussiman, Fernando de Oliveira
• Data de Publicação: 2020
• Outros Autores: Silva, Fabyano Fonseca e, Carvalho, Rachel Santos Bueno, Ventura, Ricardo Vieira, de Oliveira, Henrique Nunes [UNESP], Abreu Silva, Bárbara da Conceição, Fonseca, Mayara Gonçalves [UNESP], dos Santos, Bruna Aparecida [UNESP], Pereira, Guilherme Luis [UNESP], Eler, Joanir Pereira, Ferraz, José Bento Sterman, Mattos, Elisângela Chicaroni, Curi, Rogério Abdallah [UNESP], Balieiro, Júlio Cesar de Carvalho
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1016/j.livsci.2020.104168; http://hdl.handle.net/11449/200765
• topic: Bayesian inference; Extreme categorical problem; Gaited horse; Threshold models

Mangalarga Marchador (MM) is a very important Brazilian gaited horse breed. These animals show two gait types, differing in proportion of lateral or diagonal movements. Thus, it is possible to consider the gait type as a binary categorical trait in our statistical models. Threshold models (TM) are strongly recommended to handle binary traits for genetic evaluation purposes. However, TM are susceptible to the extreme categorical problem (ECP). ECP is usually observed due to the absence of variation within subclasses for a given systematic effect and can be avoided after handling this effect as random or by combining different systematic effects in the model. In this context, we aimed to find the most suitable systematic effect (based on goodness-of-fit and predictive ability) to be included in Bayesian threshold model for the genetic evaluation of gait type in MM horses. The dataset consisted of 1,231 gait type records and 3,172 animals in the pedigree file. Phenotypic record associated with gait type was treated as a categorical trait (MP = 0 and MB = 1). In summary, models with small complexity were benefited by smaller bias and average prediction errors. Additionally, these models showed higher heritability estimates. ECP was an important issue, and should always be approached when using threshold models for genetic evaluation.