Abstract
At first, linear multiple regression was fitted for the estimation of open days on Holstein cows using 52 parturition records. Independent variables were year (X1) and month of parturition (X2), parity (X3), body weight at conception (X4), milk production for the first 30 days (X5), body weight loss from pre-partum to 30th day post-partum (X6) and a derived variable (X7=X5•X6/1000). The multiple correlation coefficient by the equation was low.
Then, curvilinear multiple regression was fitted taking quadratic and cubic terms of independent variables (X1, X2, X5, X6, and these derivatives) into account. This time, the regression equation explained over one third of the variation in open days (adjusted R2=0.362). In that equation, the quadratic and cubic effects of year and month of parturition were significant. Milk production (X5), body weight loss (X6) and derived variables from these two had a rather complicated effect on open days. The estimated open days by the regression for all possible combinations in value of X5 and X6 showed bowl-shaped surface. When X5 and X6 took values 550 and 115kg, minimum estimate of open days was obtained as the bottom of the bowl.
