Bayesian analysis in the presence of missing explanatory variables
Hay, El Hamidi Abdel
MetadataShow full item record
Days on feed (DOF) is an essential component in the profitability of feedlot industry. Several equations have been developed to predict DOF that include multiple animal and environmental factors. More recently, molecular information (molecular breeding values) for several traits is available to add to the calculation of DOF, with the advantage that this information is readily available as soon as the animal is genotyped for a set of single nucleotide polymorphisms (SNPs). In this study, a data set of 10,209 feedlot animals were used to establish a prediction equation for DOF based on arrival weight, sex, growth, genotypes (for 200 SNP markers), and an imputed Zilmax status. Three models were compared in their prediction ability of DOF. The first model (M1) included the following effects (sex, HCW, initial weight, marbling score, backfat, ribeye area, and 36 SNPs selected out of the available 200 SNPs). The second model, M2, included all the effects in M1 plus Zilmax status (case/control) based on the predicted probabilities using 0.5 as a threshold value. The third model, M3, included all effects in M1 and the uncertain or unknown Zilmax status. Several analyses were conducted where a joint crisp and soft classification of Zilmax status was adopted.