Livestock Research for Rural Development 31 (10) 2019 LRRD Misssion Guide for preparation of papers LRRD Newsletter

Citation of this paper

Weight estimation in native crossbred Assamese goats

Andy Hopker, Jill MacKay, Naveen Pandey1, Sophie Hopker, Roopam Saikia1, Brihatrabar Pegu1, Dibyajoti Saikia1, Megan Minor, Jadumoni Goswami1, Rebecca Marsland2 and Neil Sargison

Royal (Dick) School of Veterinary Studies, University of Edinburgh, Easter Bush Veterinary Centre, Roslin, Midlothian, Scotland, UK
ahopker@exseed.ed.ac.uk
1 The Corbett Foundation, Kaziranga Office, Village Bochagaon, Kaziranga, District Golaghat, Assam 785609, India
2 School of Social and Political Sciences, University of Edinburgh, George Square, Edinburgh, Scotland

Abstract

In the Assam region, village goats serve a vital role in smallholder productions systems. Accurate bodyweight measurements are not always feasible in smallholder systems for animal husbandry, and so proxy measures may be more useful. However, bodyweight proxies are not equally informative at all life stages of the goat, and are not all equally obtainable. In this study we recorded health measures on 149 indigenous Assamese village goats including bodyweight, body length (poll to tail head), chest girth, body condition score on a 5-point thin to fat scale (BCS), and conjunctival eye colour (FAMACHA©) score. Goats in the region were thin (median score = 2) and anaemic with 82% of goats scoring a >4 on the FAMACHA scale (mean score 3.98 ± 0.69). Adult goats measured 68.0cm ±7.12 and weight 16kg ± 4.36, and kids measured 42.6cm ± 9.86 and weighed 4.19kg ±2.62. A series of linear regressions were created to predict bodyweight, and models which had clinical relevance were tested using K-folding (resampling the data k-times so all data points have both been included and excluded from the models). A quadratic regression model of girth˛, body length and pregnancy status had the lowest Root Mean Square Error (RMSE) and had significant predictive ability on bodyweight (F4,143=881.6, p<0.001). However a simpler linear model of girth and age had an acceptable RMSE and retained highly significant predictive ability on bodyweight (F2,146=758.8, p<0.001) while minimising prediction error for small goats where dosages should be more specific. We discuss the importance of selecting clinically relevant and pragmatic models in smallholder settings.

Keywords: Assam, chest girth, K-fold cross validation, smallholder farmer, small ruminants, weight estimation


Introduction

Accurate estimation of bodyweight of livestock is a prerequisite for efficient animal husbandry and correct dosing of medications. In many smallholder farming situations, the direct weighing of livestock is not practical, or the necessary equipment is unavailable, hence there is a need for alternative measurements as predictors of body weight.

The technique of measuring the chest girth of livestock in order to estimate their weight has been practiced for some time (Davis et al 1937). Initially used for dairy cattle, the method has been adapted for a variety of livestock species and breeds, however specific formulae must be used to account for variation in body morphology (Lesosky et al 2012). Chest girth measurement has been used to estimate the body weight of various types of small ruminants, including meat sheep in Italy (Sarti et al 2003); commercial sheep and goats in Nigeria (Olatunji-Akioye and Adeyemo 2009); and Central Highland and Woyto- Guji goats in Ethiopia (Zergaw et al 2017). A study across five African countries found chest girth had the highest correlation with live weight, that categorising goats by chest girth allowed different predictive models to be employed with the inclusion of other body measurements to increase accuracy, and that predictive models varied between countries within the region (Chinchilla-Vargas et al 2018).

On the Indian subcontinent various body measurements have been correlated with body weight in goats. The body weight of the Assamese Hill Goat was found to be correlated with body length, heart girth, rump length, sacral pelvic width and paunch girth (Khargharia et al 2015). Chest girth was found to be the most important predictor of body weight for Pakistani Beetal goats, with body length also being significant (Moaeen-ud-Din et al 2018). Body length, chest girth and height at withers were found to be correlated with body weight in Osmanabadi goats in Maharashtra state, India (Mule et al 2014). These studies, while thorough, do not propose straightforward practical methods to estimate the body weights of small ruminants, by linking simple body measurements to a formula for the calculation of predicted body weight. The ideal technique should be suitable for field use by veterinary clinicians, researchers, community animal health workers and livestock keepers themselves. The use of a standard tailor’s tape for measurement alongside a conversion chart, or of a tape that incorporates a formula to allow direct estimation of body weight, would broaden the scope of accessibility to the technique.

Other measurable factors could potentially confound the estimation of body weight from measurements of body proportions. Body condition score (BCS) is a standardised method used to assess the fat reserves and muscle bulk of livestock. Initially developed for cattle (Lowman et al 1976), the five point system is based upon palpation of the muscle and fat cover of the lumbar vertebrae to assess the condition of the animal, relative to its body size, irrespective of its absolute weight. The scale has been adapted for a variety of species, including sheep (Russel 1984) and goats (Villaquiran et al 2004). BCS has been found to be necessary for prediction of body weight from chest girth for some species of meat goat (Villaquiran et al 2005).

FAMACHA© scoring is a standardised method for describing the colour of conjunctival mucosa (van Wyk et al 1998). A five point scale is used where 1 is pink and 5 is severely anaemic. Targeted selected treatment based on FAMACHA© scoring has been found to be as effective as routine anthelminthic treatment of sheep in the control of haemonchosis (Leask et al 2013). Furthermore, high FAMCHA© scores have been linked to low rates of daily liveweight gain (O’Brien et al 2018).

Currently there is no validated, simple to use, technique for weight estimation of the native goats commonly found in villages of the plains of Assam. The aim of this study was to establish an appropriate method of estimation of the body weights of Assamese goats, based on measurements that are practical to use in the field by both veterinary professionals and smallholder farmers.


Materials and Methods

Village goats were examined and weighed in four different villages in the Golghat and Nagaon districts of Assam during the last week of November 2018. This region is on the flood plain of the Bramaputra River, adjacent to the Kaziranga National Park, and is subject to seasonal inundation almost every year. In these villages the main source of income is smallholder farming, with rice cultivation as the primary activity, mustard grown as a secondary crop, and vegetables grown for home consumption and sale. Livestock, mostly cattle and goats, are kept as an additional income sources, in the case of cattle to provide milk for home consumption, draft power for farming activities and calves for sale. Goats are primarily kept as a cash reserve to sell to cover family expenses such as medical bills or school fees, and it is very uncommon to slaughter a goat for household consumption. Goat keeping for dairy purposes is not commonly undertaken in the region. The goats kept in the villages of the region are generally native crossbreeds of small stature, which loosely conform to a recognisable ‘type’, whose characteristics include features of the Bengal Black Goat and the Assamese Hill Goat.

Goat reproduction is usually very loosely controlled in the region. Goats are primarily grazed on rough ground, field and road edges, and dry paddy fields. A small amount of supplementary feeding is supplied to goats, mostly in the form of collected herbage, rice straw, and vegetable waste.

Investigators went from house to house and field to field to examine goats with the permission of their keepers. Every goat in each village that was available at the time of data collection was examined. This included all goats which were tethered, presented for examination by the owner, indoors, or which the investigators were able to catch.

Chest girth, the distance around the chest incorporating the rib cage and sternum and the withers at their highest point, with the measuring tape behind the elbows, vertical and pulled tight; and body length along the spine from poll to tail head, again with the tape pulled tight; were measured using a standard tailor’s tape. Chest girth was accurately measured by a single operator, but accurate measurement of body length required both operators to hold the tailor’s tape in place. The animals were weighed using a doctor’s weigh scales (Salter, UK), with the operator standing on the scales holding the goat, before subtracting the operator’s weight from the total to ascertain the weight of the goat (Fig 1- 3.). A doctor’s weigh scale was used in preference to a hanging spring balance and sack, as a pilot study had found the former technique to be simpler, quicker, more accurate due to less movement of the goat, and to have a preferable animal welfare aesthetic. Body condition score (BCS) was assessed using the standard technique of palpating the lumbar transverse processes, and scored on a scale of 1-5, where score 1 is emaciated and score 5 is obese (Villaquiran et al 2004). FAMACHA© conjunctival eye colour score was recorded, using the standard scale of 1-5, where 1 is red- pink and 5 is white (van Wyk et al 1998). Any notable clinical observations were recorded for individual animals. Animals were assigned to either adult or goat kid age groups, where goat kids are considered animals less than approximately one year old, based on appearance. Where this distinction was not clear, there was a discussion between owner and investigators, following which the animal was assigned to one group or the other, and the animal noted as a sub-adult. It was also noted if male animals were castrated. Goats were also recorded as pregnant or non-pregnant, based on visibly obvious late term pregnancy only.

Figure 1. Body measurement and weighing in the field. Following measurement the goat was weighed by an
operator standing on the scales holding the goat in his arms, the other operator read the
scales and subtracted the weight of the human to ascertain the weight of the goat.
Note the board used to ensure the scales remained level

Figure 2. Measurement of chest girth. The tape is place behind the elbows
runureing vertically up to the withers and pulled tight

Figure 3. Measurement of body length from poll to tail head. Note that both operators
are required to hold the tape in position in field conditions
Statistical Analysis

The aim of this work was to determine a proxy of weight measurement from other body metrics in goats and provide a practical method of predicting weight in the field. Our model selection therefore sought to minimise reducible error within the prediction, while producing a model which did not trade off prediction accuracy for interpretability. However, another important consideration was that of clinical relevance. All of these needs must be achieved in a dataset of n = 149 goats.

The response variable was weight (kg), and the possible predictors were age (kid versus adult), goat sex, pregnancy status (yes/no), FAMACHA score, BCS, length (cm) and girth (cm). A range of clinically relevant models were explored as simple additive linear models, linear models with interaction terms, and quadratic models with and without interaction terms. A number of models had a high R˛ value (explaining > 91% of the variation observed in the data) and all explanatory terms in the models had alpha levels >0.01. The most challenging aspect of the analysis was model selection. To inform model selection, a range of models were explored through k-fold cross validation using the ‘rsample’ package in R (Kuhn & Wickham 2019).K-folding randomly divides the observation sets into k folds (in this case k = 10), and the first fold is considered a validation dataset. The model is fit on the remaining 9 folds, and the mean squared error is calculated for the observations in the validation fold. This is then repeated for the second fold, and so on until every individual data point has been present in the validation dataset. K-folding using k=10 is typically considered to produce an error estimate which does not have high bias or variance (James et al 2013 page 183). In practice, k-folding means that each individual data point has both been included and excluded from the model when we calculate our estimate of the error, providing a better estimate of the sample population as a whole. We then calculated the average Root Mean Square Error (RMSE) across each of the cross-validated models. The RMSE indicates the magnitude of the absolute differences between the observed value and the predicted value for a given model (Li 2012). The RMSE represents how big the ‘typical’ prediction error is for any given model, in the same units as the response variable. A smaller RMSE reflects a more accurate model. Finally, we considered the clinical relevance of each explanatory factor in the model. For example, in the early stages of model selection, pregnancy status had greater explanatory power than age, and so would supersede the inclusion of age in any given model due to the covariance between the two. This choice would make a model less able to predict the weight of very small goats, for whom the clinical implications of being over or under-dosed are more impactful. Therefore, we selected models which included age as opposed to pregnancy, despite it being technically less informative.

From this we present three models, a simple additive linear regression; a more complex polynomial regression; and a complex polynomial incorporating multiple factors. All of these models fit the data, and we present our interpretation of how these models may be used. All analysis were performed in R version 3.5.1 (Feather Spray, the R Foundation, Vienna, Austria). Results 149 goats were examined as part of the study, 67 adult females, 5 sub adult females, 3 entire adult males, 4 castrated adult males, 5 entire sub adult males, 27 female kids, 36 entire male kids, and 2 castrated male kids. 23 of the goats were observed to be in late pregnancy. The body weights, measurements, BCS and FAMACHA© scores are presented in Table 1. Goats were generally in low BCS (Fig 4.), with the overall mean BCS being 1.75 (± 0.38 SD). 82% of goats were FAMACHA© score 4 or 5 (Fig 5), indicating varying degrees of anaemia.

Table 1. Measurements and observations of goats according to age group
Mean Minimum Maximum Standard
Deviation
Kid
(n= 66)
Weight (kg) 4.2 0.5 10 2.625
Chest girth (cm) 34.1 12 46 7.771
Body length (cm) 42 20 61 9.86
BCS 1.7 1 2.5 3.887
FAMACHA© 4.1 3 5 0.6823
Adult
(n= 83)
Weight (kg) 16.0 7.5 25 4.357
Chest girth (cm) 55.8 43 68 5.273
Body length (cm) 68 55 85 7.117
BCS 1.8 1 3 0.371
FAMACHA© 3.9 2 5 0.6878


Figure 4. Body Condition Score (all goats) Figure 5. FAMACHA© score (all goats)

Five goat kids had heavy lice infestations, three had heavy flea infestations. One of these kids also had maggot wounds. One adult goat was afflicted by mange. Two kids and two adult had diarrhoea. Four kids had a marked pot-bellied appearance.

Predicting weight

The simplest model (Figure 6) had a mean RMSE of 2.10 kg across k-fold validations and predicted:

Weight (kg) = 0.47 Girth (cm) - 1.54 (Kid goats) - 10.38

A simple quadratic regression (Figure 7) had a mean RMSE of 1.57 kg and predicted:

Weight (kg) = -0.009 Girth(cm)˛ - 0.26 Girth (cm) + 2.02

The model which had the smallest RMSE (1.43 kg, Figure 8) and therefore minimised the error was more complex, incorporating length and pregnancy status, predicting:

Weight (kg) = 0.009 Girth (cm)˛- 0.48 Girth (cm)+ 0.17 Length (cm) + 1.37kg (if pregnant)+ 2.31

All terms in the three models were highly significant (Table 2). Neither BCS nor FAMACHA were found to be useful in predicting body weight in this study.

Table 2. Summary of three models predicting goat weight (kg) featuring estimates and standard errors of coefficients, t-values and associated p values
Model Estimate Std. Error t value pr(>/t/)
Intercept -10.377 1.481 -7.008 0.000
Girth (cm) 0.472 0.026 17.998 0.000
Age (Kid) -1.542 0.662 -2.328 0.021
Intercept 2.020 1.511 1.337 0.183
Girth (cm) -0.260 0.073 -3.556 0.001
Girth (cm)˛ 0.009 0.001 10.840 0.000
Intercept 2.311 1.340 1.725 0.087
Girth (cm) -0.484 0.077 -6.321 0.000
Girth (cm)˛ 0.009 0.001 12.307 0.000
Pregnant (yes) 1.368 0.340 4.022 0.000
Length (cm) 0.166 0.029 5.835 0.000


Figure 6. Linear model: Weight (kg) versus Girth (cm) + age (kid versus adult) for
149 goats, regression significant at F2,146=758.8, p<0.001


Figure 7. Simple quadratic regression: Weight (kg) versus Girth (cm)2 + Girth (cm) for
149 goats, regression significant at (F3,145 = 911.8, p <0.001


Figure 8. Complex quadratic regression: Weight (kg) versus Girth (cm)2 + Girth (cm) + Length (cm) +
Pregnant (yes/no) for 149 goats, regression significant at F4,143=881.6, p<0.001


Discussion

The use of chest girth alone as a predictor of body weight is a simple and rapid method of estimating the body weight of Assamese village goats in the field. Chest girth is quick and simple to measure, and can be undertaken accurately by operators using any measuring tape, with only brief training in the technique required. While there is some benefit in also considering body length, accurate measurement is more challenging in the field due to the natural tendency of goats to be wriggly creatures, thus a system requiring chest girth measurement only is advantageous.

The ‘best’ statistical model in a given scenario is hard to define. Here we present three highly significant models of varying complexity. As we are well aware, significance alone is a poor indicator of a model’s validity (Wasserstein & Lazar, 2016), and does not infer great confidence in the predictions. Through k-fold cross-validation, we repeatedly tested the models on withheld data to explore the difference between our predicted and observed values. The RMSE generated from this indicate the absolute fit of the model to the observed data, averaged across k-folds where each individual has not contributed to the model. From a statistical perspective, the complex polynomial model had an RMSE of 1.43kg, the simple polynomial model had an RMSE of 1.57kg and the simple linear model had an RMSE of 2.10kg. The benefit of adding more terms to the simple polynomial model is therefore an improvement in absolute fit of 0.14kg, and the benefit of the curved fit over the simple additive linear model is an improvement in absolute fit of 0.53kg. The model considering pregnancy and body length offered better overall fits, but as Figure 6 shows, the fitted line is more appropriate for kids with the simple model, and the consequences of miscalculating a dose for a smaller animal are greater.

The linear model provides a straightforward calculation which can be used in the field by a single operator with any simple measuring tape, provided the operator has basic maths skills, making this model the most suitable for an operator with no ancillary aids. The only additional judgement required is to classify the goat as kid or adult. The weight of small goat kids is accurately predicted by this model, which is important as in this group of animals small inaccuracies in prediction have a large proportional effect, which of particular importance when dosing medications. Some sub-adult goats are less accurately predicted by the model, possibly partly due to the need for the operator to make a decision on which age group to assign the animals to. Sub- adults are a key group when targeting feeding strategies for weight gain and productivity. This level of precision management is currently not commonly practiced in the region, however this highlights the need to select a model suitable for the purpose of the weigh banding activity.

The polynomial regression based on chest girth alone, is an excellent predictor of the body weight of goats of all ages and requires only one simple measurement to be taken, but necessitates more complex calculations. It is ideal for delivery using an ancillary tool, such as a specific graduated weigh tape, a conversion table or an app. Heavily pregnant goats, or those which are emaciated, may still fall outwith the models predictions, however this situation is obvious to the operator in the field, and can be readily considered.

The polynomial regression incorporating chest girth, body length and pregnancy status is theoretically the “best” predictor of bodyweight, however the difficulty of accurately measuring body length may result in incorrect estimation of weight, as well as being more time consuming to undertake in the field.

BCS did not alter the prediction of weight using these models, which aids field usage of the models, as body condition scoring is a subjective measure, and both training and experience are required to assess BCS accurately and repeatably. The likely reason for this is that as the animal gains condition the increase in fat and muscle deposited over the chest allows the model based on chest girth to account for the increase in BCS. However the girth of the bones rib cage remains unchanged, and so clinical judgement should be applied when using the modal in emaciated animals. It should be noted that the animals examined for this study were mostly in low body condition, and that this will negatively impact health and productivity.

FAMACHA© score was not useful in predicting the weight of goats in this study. This is readily explained as FAMACHA© is a useful indicator of parasitism, and of health; and as such can be considered in the prediction of weight gain, but FAMACHA© is not linked to the weight of an animal at a single point in time. The majority of the animals examined had high FAMACHA© scores, indicating anaemia, which will impact both productivity and survivability. While this study did not examine the potential reasons for this finding, parasitism is the most likely cause in this management system. Haemonchus contortus, Fasciola gigantica, Explanatum explanatum, Theileria lestoquardi, Linognatuhs stenopsis and Ctenocephalides felis are all likely to be common parasitic causes of anaemia in this landscape, but their impact, or the importance of co-infections is not currently known.

Further investigation is required to elucidate the cause of anaemia in these animals, the effect of this on health and productivity, and to explore potential mitigating strategies.

The use of a simple calculation to estimate the weight of village goats from a single measurement of chest girth provides a simple method to estimate weight, allowing for accurate dosing of medications and provides an indicator which may be used to measure productivity. The simple equation allows anyone with access to a measuring tape to predict the weight of a goat if they are able to perform a straightforward calculation. The more accurate polynomial regression equation is suitable for use for the production of specific graduated weigh tapes for these animals, which could be double sided for goat kids or adults; printed calculation tables; or the use of an app to calculate body weight from chest girth.

The production and distribution of such tools for field use through a not- for- profit mechanism would be of great benefit to farmers and community animal health workers, and would assist the drive towards rural prosperity through efficient, sustainable farming. Weigh banding should not be considered a complete solution, rather a tool, and clinical observations such as late pregnancy, emaciation, or ill health should always be considered when employing the technique.


Acknowledgements

The authors wish to thank the people of the villages of the Golaghat and Nagaon districts of Assam, without whose help and co-operation this project would not have been possible. We would also like to thank our colleagues at The Corbett Foundation and The University of Edinburgh for their help and support with this project.


Funding

This study was funded by the Royal (Dick) School of Veterinary Studies, University of Edinburgh, as part of their Indian Veterinary Education Project.


Ethical Approval

The work was carried out under ethical review by the R(D)SVS University of Edinburgh Veterinary Ethical Review Committee, approval VERC 6.18.


References

Chinchilla-Vargas J, Woodward-Greene M J, Van-Tassell C P, Wandui-Masiga C and Rothschild M F 2018 Predicting live weight of rural African goats using body measurements. Livestock Research for Rural Development. Volume 30, Article #123. Retrieved August 9, 2019, from http://www.lrrd.org/lrrd30/7/josue30123.htmlKhargharia G

Davis H P, Morgan R F, Brody S, Ragsdale A C 1937 Relation of height at withers and chest girth to live weight of dairy cattle of different breeds and ages College of Agriculture University of Nebraska Agricultural Experiment Station Research Bulletin 91 https://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1139&context=ardhistrb Accessed 30/7/19

James G, Witten D, Hastie T, and Tibshirani R 2013 An introduction to statistical learning with applications in R. (G. Casella, S. Fienberg, & I. Olkin, Series Eds.) (First Edition). New York: Springer. ISBN: 978-1-4614-7138-7 Li, Y. (2012). Root Mean Square Error. In N. Salkind (Ed.), Encyclopedia of Research Design (pp.1288–1292). Thousand Oaks: SAGE Publications. http://doi.org/10.4135/9781412961288.n392 Accessed 30/7/19

Kadirvel G, Kumar S, Doley S, Bharti P K and Das M 2015 Principle component analysis of morphological traits of Assam Hill goats in eastern Himalayan India. Journal of Animal and Plant Sciences, Vol.25(5) pp 1251- 1258

Kuhn M and Wickham H 2019 rsample: General Resampling Infrastructure. R package version 0.0.4. https://CRAN.R-project.org/package=rsample) Accessed 30/7/19

Leask R, van Wyk J A, Thompson P N and Bath G F 2013 The effect of application of the FAMACHA© system on selected production parameters in sheep Small Ruminant Research, https://www-sciencedirect-com.ezproxy.is.ed.ac.uk/science/journal/09214488/110/1 Volume 110, Issue 1, pp 1-8

Lesosky M, Dumas S, Conradie I, Handel I, Jennings A, Thumbi S, Tove P and Bronsvoort B M D 2012 A live weight–heart girth relationship for accurate dosing of east African shorthorn zebu cattle. Tropical Animal Health and Production. 45(1) pp 311–316

Li Y 2012 Root Mean Square Error. In N. Salkind (Ed.), Encyclopedia of Research Design (pp. 1288–1292). Thousand Oaks: SAGE Publications. doi:10.4135/9781412961288.n392

Lowman B G, Scott N E and Somerville S H 1976 Condition Scoring of Cattle. Bulletins (East of Scotland College of Agriculture)no. 6. https://www.worldcat.org/title/condition-scoring-of-cattle/oclc/8589575. Accessed 30/7/19.

Moaeen-Ud-Din M, Waheed H, Bilal G, Reecy, J and Khan M 2018 Estimation of Beetal goat live-weight for all types of age classes in field and farm conditions through linear body measurements. Journal of Animal Science, Vol.96 Suppl 3, pp.453-454

Mule M R, Barbind R P and Korake R L 2014 Relationship of body weight with linear. body measurement in Osmanabadi goats. Indian Journal of Animal Research, Vol.48(2), pp.155-15

O´Brien D, Schoenian S, Semler J, Gordon D and Bennett M 2018 PSVI-19 Consistency of FAMACHA© scores to fecal egg counts and gain in meat goat kids. Journal of Animal Science, Volume 96, Issue suppl 3, pp 460–461

Olatunji-Akioye A O and Adeyemo O K 2009 Liveweight and Chest Girth Correlation in Commercial Sheep and Goat Herds in Southwestern Nigeria International Journal of Morphology, Vol.27(1), pp.49-52

R Core Team 2018 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. Accessed 30/7/19

Russel A J F 1984 Body condition scoring of sheep. In Practice 6, pp 91-93

Sarti F M, Castelli L, Bogani D and Panella F 2003 The measurement of chest girth as an alternative to weight determination in the performance recording of meat sheep Italian Journal of Animal Science, Vol.2(2), pp.123-129

van Wyk J A, Bath G F and Malan F S 1998 The need for alternative methods to control nematode parasites of ruminant livestock in South Africa. World Animal Review 91, pp 30-33.

Villaquiran M, Gipson T, Merkel R, Goetsch AT and Sahlu T 2004 Body Condition Scores in Goats. Langston University Agriculture Research & Cooperative Extension. https://www.researchgate.net/publication/264889567_Body_Condition_Scores_in_Goats Accessed 30/7/19

Villaquiran M, Hart S, Gipson T A, Detweiler G, Merkel R M, Patra A and Ngwa T 2005 Prediction of meat goat body weight from heart girth measurement, body condition score and sex. Journal of Animal Science, 2005, Vol.83 Suppl 1, pp.152

Wasserstein R L and Lazar N A 2016 The ASA’s statement on p-values: context, process, and purpose. The American Statistician, 1305 (April). http://doi.org/10.1080/00031305.2016.1154108 Accessed 30/7/19

Zergaw N, Dessie T and Kebede K 2017 Using morphometric traits for live body weight estimation and multivariate analysis in Central Highland and Woyto-Guji goat breeds, Ethiopia African Journal of Agricultural Research, pp.1326-1331


Received 5 August 2019; Accepted 19 September 2019; Published 2 October 2019

Go to top