Predicted Equation in Live -Weight of Crossbred Swine Using Generalized Estimating Equation and Linear Mixed Effect Model

Abstract

The data were collected 4 times during 64 – 70 weeks of age from 173 heads of crossbred swine (Landracc x Large White x Duroc Jersey). Dependent variable is (weight, kilogram) while 6 independent variables are live-weight heart girth (L1, centimeter),  body length (L2, centimeter),  age (age, week), shoulder width (W1, centimeter) ,  hump width (W2, centimeter) and  belly width (W3, centimeter)  respectively.  Constructing predicted equation on average live - weight of crossbred swine using generalized  estimating equation (GEE)  was compared with linear mixed effect model. The results showed that in the GEE model, the intercept ,  heart girth(L1)  , body length (L2) and hump width (W2) are highly related to live- weight at 0.01 significant level when mean deviance, Pearson chi-square/df  and residual plot were considered  using  empirical standard error estimate and AR(1) correlation structure. The most  appropriate predicted equation found is Y = -191.73 + 0.4292L1 + 2.8302L2 + 0.1684W2. In linear mixed effect model with random intercept, the estimated parameters using residual maximum likelihood (REML) gave the results close to the GEE method. Equation of the mixed effect is Y_i = -191.16 + 0.4311L1 + 2.8271L2 + 0.1702W2 + _0i . Since the random variance, Var (_0i) , in each crossbred swine  was not difference ,thus this linear predicted equation  may also be used to predict the live-weight of an individual swine.