Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Ghana → DISCRIMINANT ANALYSIS AND WEIGHT ESTIMATION OF SHEEP BREEDS IN NORTHERN GHANA

University for Development Studie UDS (Ghana) 2011

DISCRIMINANT ANALYSIS AND WEIGHT ESTIMATION OF SHEEP BREEDS IN NORTHERN GHANA

Birteeb, P. T.

Titre : DISCRIMINANT ANALYSIS AND WEIGHT ESTIMATION OF SHEEP BREEDS IN NORTHERN GHANA

Auteur : Birteeb, P. T.

Université de soutenance : University for Development Studie UDS (Ghana)

Grade : Master of Science Degree in Biometry 2011

Résumé
The characterisation of the small ruminant populations in Ghana will playa major role in the maintenance of the genetic resources as the basis for future improvement in livestock production. This research was undertaken to characterise the two main breeds of sheep in Ghana morphologically by assessing variation within and between breed populations to discriminate between breeds as well as develop models for predicting liveweight of sheep at market ages. Principal Component Analysis was used to investigate the core structure of the body traits, and then Discriminant Analysis was used to discriminate and classify the sheep breeds. Regression analysis was used to develop weight prediction models. The study revealed significant differences between the morphological characteristics of the Sahelian (WALL) and the Djallonke (WAD) sheep breeds. The most discriminating traits between Sahelian and Djallonke sheep breeds were rump height, wither height, heart girth, neck girth and pin-bone width. The developed discriminant functions clearly discriminated and classified the Sahelian and Djallonke sheep into their breeds of origin, thus yielding 100%, 93.4% and 90% accurate classification for the rams, ewes and the overall sheep populations respectively. Among the simple linear models used, heart girth was the best predictor of liveweight, with R ;dj values of 84.61 - 92.36% for Sahelian sheep and 39.62 - 81.20% for Djallonke sheep. Where multiple traits were used, the best liveweight prediction model was a linear model with « : value of 95.53% although the quadratic models generally predicted liveweight more accurately than the linear models. The best prediction models were obtained for the two years old Sahelian sheep using hearth girth and body length as regressors.

Présentation

Version intégrale (29,06 Mb)

Page publiée le 10 février 2018