Вы находитесь на странице: 1из 10

Alfred George Hodges 10170757

A comparison of the accuracies of two methods of weight estimation for horses and their application to another species. Introduction People who care for animals such as veterinarians, animal management facilitators and individual animal owners require accurate weight measurements to assist in maintaining maximum health and welfare of the animal. The animals weight can be used as a general indicator of the animals health condition. Many aspects of animal health treatment rely on accurate weight measurements, for example, doses for anaesthetics and worming medications are calculated in relation to the weight of the animal (Plum C.D 1999). Livestock farmers require accurate weight measurements for use in husbandry, when determining ration requirements, as well as in industry to gauge the value of individual animals (Dhuyvetter J and Poland C 1999). Due to cost or practicality, it is not always possible for owners or carers of large animals or wild life to use weighing facilities such as electronic weighbridges, as a result a number of methods for estimating weight based on body measurements have been devised. The aim of this investigation is to determine the accuracy of two methods used to estimate horse (Equus caballus) weight and to determine if these methods can then be applied to other species, in this case the dog (Canis lupis familiaris). The null hypothesis states both methods will be equally accurate for estimating the weights of horses and dogs, in consequence to these findings the formulas based on that by Ensminger (1977) and also based on Milner and Hewitt (1969) can be applied to species other than horses to estimate weight. Method 8 Horses and 8 dogs from the University of Lincoln Stable yard and kennels were used for this study. The horses consisted of a variety of breeds all of which were geldings, ages ranged from 6 to 16 years, with the mean age of 10.6. The dogs were of various breeds, with unknown ages. The actual weight of the individual animals was measured on a weighbridge. The animals were stood stationary in their most upright positions on a flat surface for physical measurements. The body length was measured twice, both originating from the shoulders up until the hip and the buttock. The specific end points of measurement are as illustrated in figures 1and 2 for the respective animals. The heart girth was measured by running measuring tape around the circumference of both the horses and dogs, directly behind the shoulders and in front of the withers. To compare the accuracy of the estimates made with the two different methods, the mean estimated weights from each method were compared with the mean weights obtained from the weighbridge by means of a Paired Sample T-Test. The statistical package IBM SPSS was used to conduct the statistical
1

Alfred George Hodges 10170757

analysis of the data. In order to determine if the estimation methods could be applied to dogs, the significance of the probabilities was deduced on the basis that if the probability value (p) was more than 0.05 the distribution of the estimated weights was down to chance and the method used resulting in this value, was unusable to estimate the weight of the dogs as well as other species. To make the morphometric weight estimates the following formulae were used: A) Based on that by Ensminger (1977): (Heart girth cm)2 x length (shoulder to hip)cm 9500 B) Based on that by Milner and Hewitt (1969): (Heart girth cm)2 x length (shoulder to buttock)cm 12500 = weight (Kg) = weight (Kg)

Alfred George Hodges 10170757

Results Table 1 summarises the results that were produced when a Paired Sample T-Test was carried out comparing the mean weights estimated when using methods A and B for both horses and dogs, with the mean weights measured. Table 2 and 3 (Appendix 1) show the measurements taken from the horse and the calculated estimated weights. For the results obtained from the horses, the difference between the mean weight estimated using method A (M=370.2 sd=113.8) and the mean weight obtained from the weighbridge (M=583.7 sd=65.8) in conditions t(7)=5.31,p=0.001,df=7 was statistically significant .This was also the case when using method B (M=403.9 sd=102.9) in conditions t(7)=4.944, p=0.002,df=7 . The difference between the method A (M=28.4, sd=10.5) estimated mean weight of the dogs and the mean actual weight (M=24.4, sd=8.13) in conditions t (7)=-3.189,p=0.015,df=7 was found to be significant. This was also the case with method B (M=26.73, sd=8.84) when compared to the mean actual weight under the conditions t(7)=-5.866, p=0.001, df=7. When estimating the weight of the horses, both methods underestimated the mean weight, method A resulted in a larger standard deviation (113.8) and underestimated 33.6kgs less then method B (Table 2 Appendix). When applied to dogs method A also produced a larger standard deviation (3.394) compared to method B (1.105), however when estimating the mean weight of the dogs both methods overestimated, method A overestimating 700g more than method B.

Alfred George Hodges 10170757

Horses p (sig.) t df sd

Method A 0.001 5.310 7 113.780

Method B 0.002 4.940 7 102.874

Dogs Method A 0.015 -3.180 7 3.494

Method B 0.001 -5.870 7 1.105

Table 1.Results of the Paired sample t tests A correlation (c) was observed between the estimated weights and the actual weights, as the actual weight of the horse increased so did the estimated weight. For the horses method B (c=0.549) had a stronger correlation with the measured weight compared to method A (c=0.242).A much stronger correlation was exhibited between the estimated weights and the actual weights of the dogs. Method A exhibited a stronger correlation (0.961) then Method B (0.995).When plotted on a scatter graph a positive linear relationship was observed, this is illustrated in figure 3 and 4.Full results of the correlations can be found in appendix 2.
600

500

400 Estimated Weight

300

Method A Method B Linear (Method A)

200

Linear (Method B)

100

0 0 100 200 300 400 500 600 700 800 Weighed Weight

Figure.3 correlation exhibited by the estimated weights and the actual measured weights of the horses

Alfred George Hodges 10170757

50 45 40 35 Estimated Weight (Kg) 30 25 20 15 10 5 0 0 5 10 15 20 25 30 35 40 Weighed Weight (Kg) method A Method B Linear ( method A) Linear (Method B)

Figure.4 correlation exhibited by the estimated weights and the actual measured weights of the dogs Discussion Both methods used to estimate the mean body weights of the horses and the dogs, produced estimates that were significantly different (p 0.05) from the actual mean weights measured and therefore were not down to chance. For both dogs and horses method B resulted in a smaller standard deviation from the mean of the estimated weight, compared to method A, this suggests that method B was more accurate then method A. However the sample size (N=8) used for this investigation was relatively small as a result any anomalous individual variations would have had a profound effect on the mean weights measured. A study by Cattet and Obbard (2005) found that sample size affected the accuracy of estimating body mass largely through its influence on precision.

Alfred George Hodges 10170757

The standard deviations from the mean supported by the linear relationships exhibited in figure 4 shows that small discrepancies exist between the mean predicted weight and the mean actual weight of the dogs, because p 0.05 the differences in the estimated weights and the measured weights are statistically significant and not down to chance, therefore both methods of weight estimation based on that by Ensminger (1977) and that by Milner and Hewitt (1969) can be used as a reliable means to estimate the body weight for dogs as well as horses. However whether or not these methods can be applied to other species is subject to the species in question, for example Cattet et al. (1997) assessed the usefulness of morphometry to predict the body mass of polar bears (Ursus maritimus) and concluded the accuracy of this approach was too low to be useful. Species refined methods of estimation that have been altered to take into account the weight of the head have been found to produce very accurate morphometric weight estimates, for example Jansen and Jenks (2010) conducted a study to find the best method of estimating the mass of pumas (Puma concolor), their findings concluded that a model using body length (cm), head and chest circumferences (cm) explained 89% of the variation in body mass and the difference between predicted and actual body mass of pumas from the sample population was 0.40 kg 1.45. The standard deviation from the mean actual weight of horses using method A (sd=113.78) was larger than that of method B (sd=102.87) suggesting that method B was more accurate because the discrepancies between the estimated weights and the mean actual weight were smaller, as a result the null hypothesis can be rejected. The correlation cannot be used as an indication of accuracy because the p values for correlation produced when using method A (p=0.564) and method B (p=0.549) indicate the correlations are insignificant and down to chance. When taking physical measurements, in order to avoid potential sources of error, the same investigator should perform the measurements on each animal to identify specific anatomical points this ensures consistency across all animals (Henneke et al 1983) In order for more individual specific and thus more accurate methods to be established Ellis and Hollands (2002) suggested the effects of other variables such as body condition, breed and age be investigated. The difference between the standard deviations of both methods from the mean actual weight was relatively smaller with the dogs compared to that of the horses, supported by the stronger correlations observed with the estimated dog mass and the actual mass, this study suggests that morphometric weight estimates based on the equations from the work of Ensminger (1977) and that of Milner and Hewitt (1969) are more accurate when used to estimate the weight of dogs compared to horses.

6 Word count; 1366

Alfred George Hodges 10170757

Reference List Cattet M.R.L, Atkinson S.N, Polischuk S.C, Ramsay M.A (1997); Predicting Body Mass in Polar Bears: is morphometry useful? Journal of Wildlife Management 61: p1083-1090 Cattet M.R.L, Obbard M.E (2005); To Weigh or not to Weigh: conditions for the estimation of body mass by morphometry. Ursus Journal of the IBA 16(1): p102-107 Dhuyvetter J, Poland C (1999); Value Based Cattle Production. Available; http://www.ag.ndsu.edu/pubs/ansci/beef/as1163-1.htm last accessed 14th November 2011. Ellis J.M, Hollands T (2002); Use of height specific weigh tapes to estimate the bodyweight of horses. The Veterinary Record 150: p632-634 Ensminger, M.E (1977); Horses and Horsemanship (5th edition), p509. Gibbs P.G, Householder D.D (1992); Measuring and estimating weight of horses. A Method Demonstration Teaching Guide. Texas A&M University Department of Animal Science/Equine Sciences Heinrichs A.J, Rogers G.W, Cooper J.B (1992); Predicting body weight and wither height in Holstein heifers using body measurements. Journal of Dairy Science 75, p3576-3581 Henneke D.R, Potter G.D, Kreider J.L, Yeates B.F, (1983); Relationship between condition score, physical measurements and body fat percentage in mares. Equine Vet Journal 15: p371-372 Jansen B.D, Jenks J.A (2010) Estimating body mass of pumas. Wildlife Research 38(2): p147-151 Milner J, Hewitt D (1969); Weight of Horses: Improved Estimates Based on Girth and Length. Canadian Veterinary Journal 10, p314-316. Plum C.D (1999); Veterinary Drug Handbook. p853 Wheels E (2011); Measurement Chart: Available: http://eddieswheels.com/measurementorder/ last accessed 15th November 2011.

Alfred George Hodges 10170757

APPENDIX 1 Horse Sex Heart Girth (cm)


193.0 194.0 200.0 177.0 204.0 109.0 190.0 205.0

Length (cm)
Shoulder to hip 90.00 90.00 108.0 97.00 93.00 93.00 92.00 95.00 to buttock 130.0 143.0 162.0 143.0 150.0 147.0 135.0 151.0

Weight (Kg)
576.0 518.0 604.0 500.0 670.0 554.0 568.0 680.0

Estimated Weight (Kg) Method

A
385.3 389.3 496.6 388.9 334.9 127.0 381.7 458.9

B
387.4 430.6 518.4 358.4 499.4 139.7 389.9 507.7

A B C D E F G H

Gelding Gelding Gelding Gelding Gelding Gelding Gelding Gelding

Mean Weight

583.8

370.3

403.9

Table.2 Measurements taken from the individual horses and the calculated weight estimates using the method; A. Based on that by Ensminger (1977) B. Based on that by Milner and Hewitt (1969) Dog Sex Heart Girth (cm) 75.0 77.0 75.0 72.0 50.0 68.0 73.0 78.0 Length (cm)
Shoulder to hip To buttock

Weight (Kg) 63.0 74.0 64.0 62.0 42.0 59.0 67.0 76.0 26.3 32.1 27.5 23.6 7.3 20.6 24.6 33.5 24.4

A B C D E F G H

female male female female male female female female

44.0 57.0 44.0 43.0 32.0 40.0 52.0 62.0 Mean Weight

Estimated Weight (Kg) Method A B 28.4 28.4 38.8 28.4 25.6 9.2 21.3 31.9 43.4 28.4 35.1 28.8 25.7 8.4 21.8 28.6 37.0 27.7

Table.3 Measurements taken from the individual dogs and the calculated weight estimates using the same methods A and B.

Alfred George Hodges 10170757

Appendix 2 Data output from IBM SPSS for the T test carried out on the data obtained from horses Table 4. Results of the paired samples T Test for method A
Paired Samples Test Paired Differences 95% Confidence Interval of Std. Mean Pair 1 Actual estimateA 213.4250 0 Deviation 113.77937 Std. Error Mean 40.22708 the Difference Lower 118.30307 Upper 308.54693 t 5.306 df 7 Sig. (2tailed) .001

Table 4a. Correlations between method A and the actual weight


Paired Samples Correlations N Pair 1 Actual & ehorseB 8 Correlation .549 Sig. .158

Table 5. Results of the paired samples T Test for method B


Paired Samples Test Paired Differences 95% Confidence Interval of Std. Mean Pair 1 Actual ehorseB 179.8125 0 Deviation 102.87448 Std. Error Mean 36.37162 the Difference Lower 93.80728 Upper 265.81772 t 4.944 df 7 Sig. (2tailed) .002

Table 5a. Correlations between method B and the actual weight


Paired Samples Correlations N Pair 1 Actual & ehorseB 8 Correlation .549 Sig. .158

Alfred George Hodges 10170757

Data output from IBM SPSS for the T test carried on the data obtained from dogs Table.6 Results of the paired samples T Test for method A
Paired Samples Test Paired Differences 95% Confidence Interval of Std. Mean Pair 1 Actual edoga -3.93750 Deviation 3.49446 Std. Error Mean 1.23548 the Difference Lower -6.85894 Upper -1.01606 t -3.187 df 7 Sig. (2tailed) .015

Table 6a. Correlations between method A and the actual weight


Paired Samples Correlations N Pair 1 Actual & edoga 8 Correlation .961 Sig. .000

Table.7 Results of the paired samples T Test for method B


Paired Samples Test Paired Differences 95% Confidence Interval of Std. Mean Pair 1 Actual edogb -2.28750 Deviation 1.10510 Std. Error Mean .39071 the Difference Lower -3.21139 Upper -1.36361 t -5.855 df 7 Sig. (2tailed) .001

Table 7a. Correlations between method B and the actual weight


Paired Samples Correlations N Pair 1 Actual & edogb 8 Correlation .995 Sig. .000

10

Вам также может понравиться