APPLICATION OF THEORETICALLY OPTIMAL SAMPLING SCHEDULE DESIGNS FOR FIBER DIGESTION ESTIMATION IN SACCO

Three different geometrically spaced sampling schedule designs, a theoretically optimal design, and a design that included all sampling times were evaluated by comparing parameter estimates, half-life, R2, and an indicator of variance-covariance space. Alfalfa and oat hays were tested using nylon bags placed in the rumen of a fistulated, non-lactating cow, and the amount of NDF remaining was measured at specified times. Parameters were estimated from f(t, theta) = Ae(-K(t-lag)) + U, where f (t, theta) = NDF at time t (h), A = degradable NDF, U = undegradable NDF, lag = time before digestion, and K = rate constant (/h). A, U, and f(t, theta) are expressed as a fraction of DM at time 0. Estimates A and U did not fluctuate, whereas K and lag varied across designs. All R2 were over .96 and did not vary across designs. Comparison of designs that had the same number of observations showed that the indicator of the variance-covariance space was statistically similar across designs, although the optimal design was ranked best. Parameter estimates were similar when using different sampling schedule designs, but some estimates differed by 29%. The optimal design sampling schedule provided sufficient information to estimate parameters without loss of accuracy when compared with other designs.