Table 3.
Models showing how different management strategies affect milk yield of dairy cows on-farm
| Variable | Model 1: no weight1 | Model 2: weight by N2 | Model 3: remove missing3 | Model 4: replace missing4 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Est | SE | P-values | Est | SE | P-values | Est | SE | P-values | Est | SE | P-values | |
| Intercept | 7.91 | 1.31 | <0.001 | 6.85 | 0.86 | <0.001 | 7.57 | 1.46 | <0.001 | 7.79 | 1.41 | <0.001 |
| Days in milk | −0.02 | 0.00 | <0.001 | |||||||||
| Calf suckling | −2.06 | 0.60 | <0.001 | |||||||||
| Trial length | −0.01 | 0.00 | 0.004 | −0.01 | 0.00 | 0.056 | −0.01 | 0.00 | 0.020 | |||
| Improved cattle breeds | 2.79 | 0.54 | <0.001 | 2.11 | 0.36 | <0.001 | 2.43 | 0.63 | <0.001 | 2.53 | 0.52 | <0.001 |
| Improved feeding | 0.40 | 0.73 | <0.001 | 1.25 | 0.46 | 0.007 | 1.89 | 0.34 | <0.001 | 0.06 | 0.84 | <0.001 |
| Maximum temperature | −0.13 | 0.04 | <0.001 | −0.12 | 0.05 | <0.001 | −0.11 | 0.04 | 0.010 | |||
| Imp. catt. breeds × Imp. feeding5 | 1.48 | 0.78 | 0.06 | 1.94 | 0.88 | 0.03 | ||||||
| Fit statistics | ||||||||||||
| N | 102 | 119 | 92 | 102 | ||||||||
| Observed mean | 6.37 | 6.97 | 6.52 | 6.37 | ||||||||
| Predicted mean | 6.37 | 9.95 | 6.45 | 6.31 | ||||||||
| RMSE, % mean | 15.82 | 17.2 | 16.4 | 16.20 | ||||||||
| Mean bias, % MSE | 0.00 | 0.02 | 0.42 | 0.38 | ||||||||
| Slope bias, % MSE | 0.48 | 1.07 | 0.72 | 0.03 | ||||||||
| RSR | 0.81 | 1.08 | 0.83 | 0.80 | ||||||||
| CCC | 0.93 | 0.92 | 0.93 | 0.93 | ||||||||
| 2.19 | 3.34 | 2.33 | 2.19 | |||||||||
| 1.13 | 5.23 | 2.56 | 2.49 | |||||||||
| AICc | 390 | 549 | 368 | 399 | ||||||||
Est = estimate; N = number of daily milk yield observations considered to fit the model; MSE = mean squared error; RSR = root mean squared error divided by population SD; CCC = concordance correlation coefficient;
= square root of the estimated study variance;
= square root of the residual variance; AICc = corrected Akaike information criterion.
Fitted using no weighting.
Fitted using weighting based on the number of observations for each management practice.
Using weighting based on 1/SE, and all observation without the SE excluded.
Using weighting based on 1/SE with mean SE used for all observations with missing SE.
Interaction effect between improved cattle breeds and improved feeding.