General linear model procedure

Analyses were performed using the general linear model procedure (SAS Institute, Cary, NC, USA). Data on Chemical composition and characteristics data of fresh and ensiled TMR were subjected to one-way analysis of variance (ANOVA) with treatment as factor. In aerobic conditions, the data on chemical composition of silages were 4 (treatments)×4 (deterioration periods)×3 (replicates) = 48 observations corresponding to each variable and were analyzed in a repeated measures analysis of variance using the PROC GLM. In General Linear Model, seven various covariance structures (CS, UN, HF, AR, ARH, ANTE) were applied. AIC and AICC criteria were used to determine the most appropriate covariance pattern for fitting data, it was determined that unstructured (UN) covariance structure gave the best fit to data set. The triplicate samples were considered as replicates, and treatments and deterioration periods were considered to be between- and within-subjects factor, respectively. All unstructured covariance matrix of the data were meet the assumption of sphericity. Statistical difference between means was determined by Tukey’s multiple comparison. Differences were considered significant when probability was less than 0.05.

Chemical composition data were analysed in a randomized complete block design with 5 replications arranged in a 2 × 3 factorial using software package of SAS Institute Inc., (2003) , version 9.1.3. In vitro and in sacco degradability data were analysed using the NEWAY computer programme for estimation of degradation constants (Osuji et al., 1993 ). The analysis of variance was carried out on the chemical constituencies and on DM disappearance coefficients a, b, c, a + b and P of ICEAP 00557, ICEAP 01514 and CIMMYT 100/01 using the SAS program General Linear Model Procedure (SAS Institute Inc, 2003 ). Differences between treatment means were assessed by Least Significant Difference. The following model was used:
$Y_{ijkl}=\mu +T_i+V_j+D_k+{\left(VDT\right)}_{ijk}+E_{ijkl}$ where Y_ijkl = observation of the dependent variable; μ = fixed effect of population mean for the variable; T_i = effect of incubation time h (0, 4, 8, 12, 24, 30, 48); V_j = effect of variety (j = 3; ICEAP 00557, ICEAP 01514 and CIMMYT 100/01); D_k = effect of drying method (k = Oven-drying and shade-drying); VDT_ijk = effect of interaction among variety at level j, drying method at level k and incubation time at level l; E_ijk = the random error associated with observation ijkl.

Outliers were tested using the UNIVARIATE procedure (SAS Inst. Inc., Cary, NC, USA). In Exp. 1, the data for the ATTD of GE and the DE and ME content were analyzed as a completely randomized design using the general linear model procedure (SAS Inst. Inc., USA), with pig as the experimental unit. In Exp. 2, data were analyzed using the MIXED procedure (SAS Inst. Inc., USA). The statistical model for the digestibility values had diet as a fixed effect and period and pig as random effects. Multiple comparisons were conducted using Tukey’s method. In all analyses, the differences were considered significant if p<0.05.