Milk fatty acids as covariates in multiple regression analysis is a robust approach to model the decrease in milk fat concentration and yield in small ruminants

Milk fat depression (MFD) syndrome has been associated with the antilipogenic effects of trans‐10 fatty acids (FA), such as t10, c12‐CLA (CLA) and t10‐18:1 (T10). However, these FA alone cannot completely explain the changes in milk fat in small ruminants. Thus, the aim of this study was to use multiple regression analysis to evaluate other FA that may be related to shifts in milk fat, as well as to improve model accuracy when different milk FA are used as covariates in the models. Previously published data were used in multiple regression analysis for goats (n = 106) and ewes (n = 68). Body weight (BW), vaccenic acid (t11‐18:1), both trans‐10 FA and the major milk FA were tested as covariates to model four response variables associated with MFD: fat concentration (FC), percentage change in milk fat concentration (CFC; %), fat yield (FY; g/d) and percentage change in milk fat yield (CFY; %). All four multiple regression models were significant for both species. When compared with simple regression models, all multiple regression models improved accuracy when estimating MFD. The improvements in model accuracy (lower RMSE) for FC, CFC, FY and CFY were 60.6%, 43.3%, 35.6% and 44.4% for ewes, and 52.1%, 60.1%, 33.6% and 14.9% for goats respectively. Linolenic acid and t11‐18:1 were covariates in all models for goats, and palmitic acid and CLA were covariates in all ewe models. These FA should be investigated regarding their direct effect on gene expression associated with milk fat metabolism in the mammary gland of small ruminants. Multiple regression analysis is the most robust approach to account for the variation of milk fat and yield in goats and ewes.