Simple and robust model to estimate live weight of Ethiopian Menz sheep

Wamatu, J., Alkhtib, A. ORCID: 0000-0002-3381-0304 and Rischkowsky, B., 2019. Simple and robust model to estimate live weight of Ethiopian Menz sheep. Animal Production Science, 59 (12), pp. 2265-2272. ISSN 1836-0939

[img]
Preview
Text
13513_a1421_Alkhtib.pdf - Post-print

Download (470kB) | Preview

Abstract

Heart girth (HG) bands have been predominantly used in Ethiopia by smallholder farmers, traders and extension workers to estimate live weight (LW) of livestock. They are produced using recommended and published predictive models from Ethiopia. More recently, some farmers and traders have abandoned the bands due to perceived inaccuracy of LW estimation and reverted to eye ball estimations. This study generated a novel algorithm using multiple criteria to develop a robust predictive model for LW estimation of Ethiopian Menz sheep using HG. Subsequently, recommended models currently in use in Ethiopia were evaluated for accuracy in predicting LW using data of this study. Live weight and HG of 420 Menz sheep were measured. Simple linear model (SLM), Box-Cox (SLM with LW0.75), quadratic and allometric models were used to describe the relationship between LW and HG. Algorithms used to validate the models included data exploration, model construction and model redeployment. Results revealed that all models had similar R2 (≈0.82). All models fitted the criteria of residuals analysis and robustness against extreme values. However, only Box-Cox was robust against data redeployment with 95th percentile of prediction error (PE) less than 10%. Accordingly, a Box-Cox model (LW0.75 = -9.71 + 0.289(HG)) is robust and can be used to accurately predict LW of Menz sheep. The 95th percentile of PE of existing, recommended models was higher than 10, thus they cannot be recommended to accurately predict LW of Menz sheep. This study concludes that an approach based on regressing LW on HG then selecting models with highest R2 is inadequate to generate accurate and robust prediction models. This highlights the importance of model redeployment to generate accurate prediction models. Calibrated HG bands are suitable alternatives to weighing scales in rural areas of Ethiopia because they are cheaper and not subject to maintenance. Thus, their accuracy and robustness in estimation of LW is vital for sustainable use.

Item Type: Journal article
Publication Title: Animal Production Science
Creators: Wamatu, J., Alkhtib, A. and Rischkowsky, B.
Publisher: CSIRO Publishing
Date: 2019
Volume: 59
Number: 12
ISSN: 1836-0939
Identifiers:
NumberType
10.1071/AN18006DOI
Divisions: Schools > School of Animal, Rural and Environmental Sciences
Depositing User: Jill Tomkinson
Date Added: 11 Mar 2019 11:49
Last Modified: 21 Feb 2020 10:12
URI: http://irep.ntu.ac.uk/id/eprint/35960

Actions (login required)

Edit View Edit View

Views

Views per month over past year

Downloads

Downloads per month over past year