Prediction of live weight from linear body measurements of indigenous goats of Swaziland

M Matsebula, E Bhebhe, J F Mupangwa and B J Dlamini

University of Swaziland Department of Animal Science, Faculty of Agriculture,
PO Luyengo, Swaziland
ebhebhe@uniswa.sz

Abstract

The study was conducted to identify linear body measurements (neck, heart girth, and hip bone height) which could be used in building an effective prediction model for liveweight of goats. In addition to linear body measurements, the study also investigated whether sex and age had a significant contribution in the accuracy of the prediction model. The desire to development a weight band for goats was driven by realization that resource poor goat farmers are not in a position to purchase their own weighing scales and that their inability to determine the weight of their animals compromised both management and marketing strategies. Data were collected from 300 goats of different ages and sex. Liveweight for each goat was determined using a hanging balance and linear measurements were determined using a tailor’s tape. Stepwise regression analysis determined that heart girth and hip bone length were the major significant contributors to the liveweight prediction equation.

The coefficients of determination (R²) for the exponential models for heart girth and hip bone height against liveweight were 0.92 and 0.90 respectively. Model testing using an independent group of 100 goats produced correlation coefficient values of 0.89 and 0.86 between predicted liveweights and actual liveweights for the heart girth and hip height models respectively.   It was concluded that either heart girth or hip bone height could be used with acceptable accuracy as single independent model predictors of liveweight of local goats. It was noted that if one is working alone determination of hip bone height was easier compared to obtaining a heart girth measurement.

Key words: exponential regression, heart girth, hip bone, prediction equation

Introduction

Goats are multi-functional animals and play a significant role in the economy and nutrition of landless, small and marginal farmers (Khan et al 2006 XE "Khan et al 2006" ). In Swaziland goat rearing is an enterprise which is practiced by a large section of the rural population. Goats can efficiently survive on available shrubs and trees in the harsh low fertility environment where hardly any crops can be successfully grown. They contribute to the livestock industry in terms of meat, skin and hair. According to Khan et al (2006) in comparison with other domestic animals, goats are the victim of prejudice and neglect, but they have nevertheless fulfilled a most useful task of supplying a part of the humane population with milk, meat, hair, leather and other products.

According to the Swaziland livestock census figures for 2007 (MOAC 2007) Swaziland has a goat population of about half a million goats. A large number of these goats are raised at subsistence level by resource poor farmers on Swazi Nation land. According to de Villiers et al (2009) income derived from goat production is a major contributor to the livelihoods of rural people. In order to get an income from goats, they need to be managed properly. Because of their financial situation these farmers are unable to procure weighing scales for determining the weight of their animals. As a result farmers rely on estimation of liveweights for various purposes including feeding, when to breed, determination of dosages of various medications and vaccination. Visual determination of the weight of animals is often faced by errors exacerbated by among other things, using the same estimate for more than one breed of a particular species. Body structure can be deceptive when estimating weight (Otoikhian et al 2008). The net result of reliance on estimated liveweights is inefficiencies in both the management of animals and inconsistencies in prices of goats which all work against the farmer.

A weight band is a measuring tape whose graduation has a correlation between liveweight (measured in kg) and linear measurement (measured in cm). The idea of a weight band is neither new nor novel with weight bands having been developed and in use for dairy cows (AFBI 2011; Dingwell et al 2006), beef cattle (Nesamvuni et al 2000; Bozkurt 2006; Machilaa 2008; Abdelhadi and Babiker 2009), pigs (Machebe et al 2010; Mutua et al 2011) sheep (Fouire et al 2001; Topal et al 2003; Topal and Macit 2004) and goats (Nesamvuni et al 2000; Slippers 2000; Nsoso et al 2003; Thiruvenkadan 2005; Khan et al 2006; de Villiers et al 2009; Yakubu et al 2011).  The weight band is a simple yet practical and effective tool for livestock farmers without access to an appropriate livestock scale. Availability of a weight band and farmer training on how to use it would assist resource poor farmers in implementing good animal husbandry practices.

The knowledge of livestock weight assessment remains the backbone on which all animal production management practices are hinged. Apart from avoiding the errors of visual determination of animal weights, small scale animal farmers need a reasonable and simple skill in estimating liveweight of goats when a weighbridge cannot be accessed. Linear body measurements can be used to assess growth rate, feed utilization and carcass characteristics of farm animals (Brown et al 1973). According to Birteebi and Ozoje (2012) , when producers and buyers of livestock are able to relate animal measurements to growth characteristics, optimum production and value-based trading systems will be realized. This would insure that farmers get value for their stock rather than the middlemen and or livestock processors making a lot more profit than the rural farmer (Afolayan et al 2006.

Increasing the genetic potential for meat production of goat breeds requires selection for increased size and liveweight. Proper size and weight measurements are often difficult in villages due to lack of weighing scales. Although as noted earlier, a wealth of information is available on linear body measurements as predictors of liveweight in goats, no literature of this nature for Swazi indigenous goats could be located. Needless to say given that growth, final body weight, confirmation etc vary between goat breeds, it would be potentially misleading to assume that equations developed for a different breed or strain of goat under different conditions would be as effective for Swazi indigenous goats. The objective of this study was to develop and test for models for predicting liveweights in Swazi indigenous goats using linear body measurements.

Materials and Methods

Research design and procedure

A total of 300 Swazi indigenous type goats from 35 different households were used in the study. Data was collected by way of homestead visits over a period of one and half months. To minimize on potential effects of measurement time on the liveweights, all measurements were done in the morning (Before 12 noon).

The data collected included liveweight (LW), sex, age based on dentition (AG), hip bone height (HBH), heart girth (HG) and neck circumference (NC) for each of the 300 goats. Liveweights were measured with the aid of a specially designed weighing harness and a hanging balance while linear measurements were obtained using a measuring tape.

Statistical analysis

The data were analysed using STATISTIX^© (2008). The initial full model fitted was:

У_ij= b₀+ β₁(c_ij)+β₂(c_ij)+β₃(c_ij)+β₄(c_ij)+e_ij  Where:

У_ij = the weight of the j^th individual (j=1….300) with different combinations of i (i=1….5)

b₀ = population constant common to all observations (intercept)

b_i= the regression coefficient for the i^th independent variable; i=1,….,5; 1= fixed effect of sex of goat; (male, female); 2=fixed effect of age (mature, immature); 3=effect of neck circumference; 4=effect of heart girth and 5= effect of hip bone height

Model assumes: e_ij ~ N (0, δ²)

Identification of independent variables that significantly (P<0.05) contributed to the prediction model was achieved through a stepwise regression procedure in STATISTIX^© (2008) XE "STATISTIX^© (2008)"  with the P-levels for both entry and exit from the model set at 0.05. Several model forms were tried out but based on residual scatter gram analysis, the final models adopted as best fitting the data were of the exponential form.

Validation of the prediction model

The generated models were tested by a “blind fold procedure” using flocks of goats which were not used during the development of the prediction equations but fitting the Swazi indigenous goat type. One hundred goats were used in testing the fidelity of the prediction equations. The liveweight and linear body measurements (only of the independent variables included in the final equations) were determined for the 100 goats. Using the proposed prediction equations, the linear body measurements of the 100 goats were used to generate a set of expected liveweights. The fidelity of the prediction models and hence the validity of using future weight bands developed using this information was assessed by determining the correlation between the expected liveweights and the actual liveweights for the 100 goats.

Results

Table 1 is a summary of the stepwise regression output for determining the significance and hence order of entry of independent variables into the model and elimination of those that did not meet the P<0.05 significance level.

Heart girth was the first independent variable to enter the model (P<0.0001), followed by hip bone height (P<0.0001) and sex entered the model last (P=0.014). Variables neck circumference and age group were did not contribute significantly (P>0.05) to model prediction and were hence dropped from the model.

Although sex was significant (P=0.014), it was however noted that its exclusion in a reduced model had little impact on the model R² value and hence the prediction power of the prediction model. The R² value of the full model as selected by the stepwise procedure (LW=SEX+HG+HBH) was 0.8601where as a reduced model generated by excluding sex from the model (LW=HG+HBH) had an R² value of 0.8572, a reduction in the R² value by a mere 0.003 points. Since the primary objective of the study was development of a simple and rapid yet reasonably accurate method for predicting liveweight, sex was eliminated from the model solely for convenience.

When fitting exponential models with the remaining independent variables (HG and HBH), the best single predictor of liveweight was the HG with R²= 0.92 (P<0.01) followed by the HBH with R²= 0.90 (P<0.01).

Model validation produced correlation coefficient values between the actual liveweights and predicted liveweights of 0.89 and 0.86 for heart girth and hip bone height respectively. 

Figure 1. Prediction equation for live weight (kg) with heart girth (cm) as a single independent variable in the model	Figure 2. Distribution of residuals for the prediction model for live weight (kg) of Swazi indigenous goats using heart girth (cm)

Figure 3. Prediction equation for live weight (kg) with hip bone height (cm) as a single independent variable in the model	Figure 4. Distribution of residuals for the prediction model for live weight (kg) of Swazi indigenous goats using hip bone height (cm)

Tables 2 and 3 respectively present the predicted liveweights given heart girth or hip bone height measurements. Although the ultimate goal is to provide “direct read-off” weight bands to farmers, for the time being all that a farmer needs to predict the liveweight of their goat stock is one of the two tables and a simple tailor’s measuring tape.

Discussion

In order to practice good animal husbandry, the measurement of live body weight is totally essential for breeding, nutrition and health management. Findings of this investigation are in agreement with literature pointing out that HG is a better predictor of liveweight closely followed hip bone height (de Villiers et al 2009; Yakubu et al 2011). The R² value of 0.92 for the exponential model with heart girth as the sole independent variable in the model reported in this study is in close agreement with de Villiers et al (2009), who used nonlinear regression models on data collected from 1202 Boer goats and reported an R² value of 89.4 for a heart girth based prediction model.
 

In the current study both age and neck circumference did not significantly contribute (P>0.05) to the prediction equation and hence were excluded from the final models.  While sex was significant (P=0.014), its impact on the accuracy of the prediction model was very small adding only 0.003 to the model’s R² value and hence it was decided to drop it from the model to enhance user friendliness of the procedure.

Predicted liveweights based on heart girth measurements had a 0.89 correlation to the actual liveweights while predicted liveweights obtained using a hip height model gave a correlation coefficient of 0.86 to the actual liveweights. The high correlation values between predicted and actual liveweight measurements give credence to the validity of using these models for predicting liveweights of Swazi goats.

Conclusion

• Of the linear body measurements investigated in this study, heart girth was the best predictor of liveweight closely followed by hipbone height. Age based on dentition and neck circumference did not significantly contribute to the prediction models. The final models settled for were of the exponential form, this was based on choice of a model that best described the relationship of both heart girth and hip bone height to liveweight.

References

Abdelhadi O M A and Babiker S A 2009 Prediction of zebu cattle live weight using live animal measurements. Livestock Research for Rural Development. Volume 21, Article #133. Retrieved June 5, 2013, from http://www.lrrd.org/lrrd21/8/abde21133.htm

AFBI 2011 AFBI/CAFRE Survey - Breeding dairy heifers - When to breed? http://www.afbini.gov.uk/index/news/news-releases/news-releases-archive-2011.htm?newsid=18965

Afolayan RA, Adeyinka IA and Lakpini CAM 2006 The estimation of live weight from body measurements inYankasa sheep. Czech Journal of Animal Science 51: 343-348.

Birteebi P and Ozoje M O 2012 Prediction of live body weight from linear body measurements of West African long-legged and West African dwarf sheep in Northern Ghana. Online Journal of Animal and Feed Research .Volume 2, Issue 5: 427-434.

Bozkurt Y 2006 Prediction of body weight from body size measurements in Brown Swiss feedlot cattle fed under small-scale farming conditions. Journal of  Applied Animal Research 29: 29-32.

Brown J E, Brown C J and Butts W T 1973 Evaluating relationships among immature measures of size, shape and performance of beef bulls 1; principal component as measures of size and shape in young Hereford and Angus bulls. Journal of Animal Science 36: 1010-1020.

de Villiers J F, Gcumisa S T and Gumede S A 2009 Estimation of live body weight from the heart girth measurement in KwaZulu-Natal goats. Applied Animal Husbandry & Rural Development 2009, Volume 2. Retrived June 5, 2013, from http://www.mdukatshani.com/resources/Weight%20band%20English.pdf.

Dingwell R T, Wallace M M ,McLaren L C F and Leslie K K E 2006 An evaluation of two indirect methods of estimating body weight in Holstein calves and heifers. Journal of Dairy Science 89: 3992-3998.

Fouire P J, Neser F W C, Oliver J J and Van der Westhizen C 2002 Relationship between production performance, visual appraisal and body measurements of young Dorper rams. South African Journal of Animal Science 32: 256-62.

Khan H, Muhammad F, Ahmad R, Nawaz G, Rahimullah A and Zubair M 2006 Relationship of body weight with linear body measurements in goats. Journal of Agricultural and Biological Science 1: 51-54.

Machebe N S and Ezekwe A G 2010 Predicting body weight of growing-finishing gilts raised in the tropics using linear body measurements. Asian Journal of experimental biological sciences 1:162-165.

Machilaa N, Fèvreb E M, Maudlina I and Eisler M C 2008 Farmer estimation of live bodyweight of cattle: Implications for veterinary drug dosing in East Africa. Retrieved June 5, 2013, from http://www.sciencedirect.com/science/article/pii/S0167587708001165.

Ministry of Agriculture and Co-operatives (MOAC) 1999 Swaziland Livestock Census. Ministry of Agriculture and Co-operatives, Data Processing Unit, Mbabane, Swaziland.

Mutua F K, Dewey C E, Arimi S M, Schelling E and Ogara W O 2011 Prediction of live body weight using length and girth measurements for pigs in rural Western Kenya. Journal of Swine Health and Production 19(1):26-33. Retrieved June 5, 2013, from http://www.aasv.org/shap/issues/v19n1/v19n1p26.html

Nesamvuni A E, Mulaudzi J, Ramanyimi N D and Taylor G J 2000 Estimation of body weight in Nguni type cattle under communal management conditions. South African Journal of Animal Sciences 30(1): 97-98.

Otoikhian C S O, Otoikhian A M, Akporhuarho O P, Oyefia V E and Isidahomen C E 2008 Body measurement parameters as a function of assessing body weight in goats under on-farm research environment. African Journal of General Agriculture 4(3): 135-140. Retrieved February 5, 2013, from http://www.asopah.org

Statistix^© 2008 Statistix for windows. Tallahassee, Analytical Software, FL 32317 USA.

Yakubu A, Ladokun A O and Adua M M 2011 Bioprediction of Body Weight from Zoometrical Traits of Non-descript Goats using Linear and Non-Linear Models in North Central Nigeria. Livestock Research for Rural Development. Volume 23, Article #130. Retrieved May 9, 2013, from http://www.lrrd.org/lrrd23/6/yaku23130.htm

Received 12 June 2013; Accepted 12 June 2013; Published 1 August 2013

Go to top