Statistical Methods for Improving Sensitivity in Dairy Cattle Feeding Experiments

Abstract

This study compares four methods of sta- tistically analyzing performance data from a dairy cattle feeding experinmnt. The ba- sic design (modified switch-back) involved a pre-experimental (Period I), an experi- mental (Period II), and a postexperimental (Period III) period. Periods I and III were standardization, or control periods, during which all cows were fed the same. Thirty-sLx Holsteins were used to determine the effect of six treatments (2 X 3 fac- torial) on milk and solids-corrected milk yield, milk fat and solids-not-fat percent- age, dry matter and estimated net energy intake, and estimated net energy/solids- corrected milk. The four methods of analy- sis ntilized different data as follows: (A) average performance in Period II, (B) average performance in Period II adjusted by covariance for average performance in Period I, (C) average performance in Pe- riod II adjusted by covariance for both average performance in Period I and body weight at the end of Period I and (D) difference in average performance in Pe- riod II relative to that in Periods I and III. The difference method of analysis yielded the smallest error variances and coefficients of variation for each of the traits studied. Compared to Method A, Method D reduced the error variance for milk yield by 94%, estimated net energy intake by 65%, and fat percentage by 52%. Use of covariance adjustment for Period I performance also reduced error variances to almost the same extent without requiring data from Period III. The inclusion of body weight as a second eovariable did not result in any ap- preciable reduction in error variance com- pared to pre-experimental performance as a single covariable. Consequently, from the standpoint of sensitivity of dairy cattle feeding experiments, the difference method would be the preferred method among those compared; however, analysis using pre-