Development and evaluation of feeding-period average dry matter intake prediction equations from a commercial feedlot database

J. P. McMeniman, L. O. Tedeschi, P. J. Defoor, M. L. Galyean

Research output: Contribution to journalArticlepeer-review

7 Scopus citations

Abstract

Intake prediction equations were developed from a commercial feedlot database consisting of 3,363 pen means collected from 3 feedlots over a 4-yr period. The objective was to predict feeding-period average DMI from variables known at the start of the feeding period, including initial BW, sex, and average DMI from d 8 to 28 of the feeding period. Equations were evaluated within the same database using bootstrapping and cross-validation techniques. Bootstrapping evaluations of equations that included initial BW and sex explained approximately 57.0 to 73.2% of the variation in observed DMI with 90.5 to 96.6% accuracy, but DMI was overpredicted by 0.21 kg/d. Accuracy and precision improved with addition to the models of average DMI from d 8 to 28 of the feeding period; these models accounted for 68.0 to 83.0% of variation in observed DMI with model accuracy between 95.2 and 99.5% and overprediction of DMI by 0.05 kg/d. Cross-validation of developed equations confirmed the robust nature of chosen variables and the decrease in prediction error as sex and average DMI from d 8 to 28 were added to models based on initial shrunk BW (SBW). Bootstrapping and cross-validation evaluations were also conducted on the NRC (1996) NEm-based intake prediction equations with and without the NRC-recommended 4% decrease in predicted DMI and a 12% increase in dietary NEm concentration associated with the feeding of monensin. The average metabolic SBW used in these equations was computed from initial SBW and a predicted final SBW developed from the pen database. When these evaluations were performed, model precision decreased compared with simpler equations that included initial BW and sex, but using both adjustments for use of monensin improved the accuracy of the NRC equation. Nonetheless, the systematic bias proportion of the mean square error of prediction for bootstrapping analyses increased from 4.6 to 11.1% in the model with no adjustments to 16.3 to 38.9% in the model with both monensin adjustments, demonstrating these adjustments are less than optimal. Overall, variables that are typically available to feedlot managers when cattle are started on feed (e.g., initial BW, sex, and an early indication of DMI) were important predictors of feeding-period average DMI by pens of cattle.

Original languageEnglish
Pages (from-to)3009-3017
Number of pages9
JournalJournal of animal science
Volume88
Issue number9
DOIs
StatePublished - 2010

Keywords

  • Dry matter intake
  • Feedlot cattle
  • Intake prediction equation

Fingerprint Dive into the research topics of 'Development and evaluation of feeding-period average dry matter intake prediction equations from a commercial feedlot database'. Together they form a unique fingerprint.

Cite this