应用ML和REML法监测基于Wood模型奶牛泌乳曲线形状的研究

Study on Detecting Shape of Lactation Curve of Cattle Based on Wood Model by ML and REML

利用最大似然法(ML)估计了内蒙古三河牛泌乳曲线基于不完全伽马函数(Wood)的不同统计模型,对各模型的差异用似然比(LR)进行检验。结果表明,模型Ⅲ显著优于模型Ⅱ和模型Ⅰ(P<0.01)。采用限制性最大似然法(REML)对模型Ⅲ估计得出,泌乳初期奶量在不同胎次(pari)、产犊年度(year)、产犊季节(seas)、场(herd)下具有显著差异(P<0.05);测定日产奶量的变化率随测定日(t)、测定日自然对数(t′)的推移,以及胎次与测定日(t)及其自然对数(t′)交互作用的不同差异极显著(P<0.01),泌乳初期奶量与t′和t的斜率系数之间存在显著的正或负相关(P<0.01)。

Estimated the different statistical models of Inner Mongolia Sanhe cattle＇s lactating curve which was based on mis-classification γ function（Wood model） by maximum likelihood method（ML）,then test the differences by likelihood ratio（LR）.As a result,model Ⅲ was significantly better than the model Ⅱ and Ⅰ（P0.01）.Estimated the model Ⅲ by REML,there were siginicantly differences of milk yield in initial stage under different pari,year,seas and herd（P0.05）.The growth rate of diraly milk yield had significantly difference with the test days（t）,natural logarithm of the test days（t′）,interaction of the pari and test days,interaction of the pari and natural logarithm of the test days.Milk yield in initial stage has positive or negative correlation with the slope rate coefficient of t and t′（P0.01）.