Accuracy Of Alternative
Equine Weight Estimation Models Utilizing New Variables
© 2002 by Alyssa Hapgood
All Rights Reserved
Mount St. Mary High School, Oklahoma City
The purpose of this study was to investigate the development of new equine weight estimation models while analyzing the comparative accuracy of numerous existing methods. Data was collected on a new sample of 77 horses ranging in weight from 300 to 2,010 lbs., and in height from 10.5 to 18.25 hands. From this year’s research data, five new weight estimation models were developed utilizing Multiple Regression Analysis (MRA). Heart Girth (Girth1), Umbilical Girth (Girth2), Height, Length from the tuber ischium to the point of the shoulder (Length1), Length from the tuber ischium to the olecranon (Length2), Age, Breed, and Gender were recorded for each horse in the research sample. Additionally, a visual estimate of each horse’s weight was made prior to obtaining its actual weight on a digital scale.
Weight estimates for each horse were also calculated using three commercially available girth weight tapes. These were the Purina, Nutrena, and Sure Measure weight tapes. Additionally weight estimates were calculated using the Carroll/Huntington formula, a traditional formula, the Ensminger Formula, the Jones, et. al formula, and formulae developed in previous phases of this project. After the data collection phase was completed, comparisons were made between the equine weight estimates using the girth weight tapes, visual estimates, the other existing formulae, and the new regression models developed.
Overall accuracy comparisons were made by comparing the adjusted R square statistics for each estimation technique/model, as well as the average percent accuracy. The new estimation models built during the project included Girth1 (heart girth), Girth2 (umbilical girth), Height, Length1 (length from the tuber ischium to the point of the shoulder), and Length2 (length from the tuber ischium to the olecranon) as independent variables and actual weight as the dependent variable.
The new models constructed through this year’s research project indicate mixed results. Their predictive accuracy was less than those developed during last year’s research from the variables Girth1 (heart girth), Length1 (length from the tuber ischium to the point of the shoulder) and Height. This year’s best model was created using Girth1 (heart girth), Length2 (from the tuber ischium to the olecranon) and Height, but its predictive accuracy was still slightly less than last year’s models.
It is critical for horse owners, veterinarians, trainers, and others involved in the care and management of the horse to accurately establish equine weight.1 An accurate weight is required for medication dosing, nutrition requirements, training purposes, and general health monitoring of the horse.2
Accurate equine weight is vital for administration of anthelmintics and anesthesia as well as other medications.3 Underdosing of anthelmintics may possibly result in parasite resistance and chronic equine health problems, while underdosing of anesthesia results in unnecessary surgical complications.4
The most accurate method for determining equine weight is a digital scale. Due to the high cost of large animal scales, the average horse owner or trainer is not likely to own or have access to a scale.5 Furthermore, many equine veterinarians respond to calls in the field, and do not have access to a scale when assessing, treating, and administering medications to horses in their care, as opposed to a clinical setting where availability of scales would be more likely.6
The most commonly used method for determining equine weight is visual estimation.7 A 1990 study by Asquith et al evaluated the proficiency of horsemen and veterinarians in estimating the weight of various equines using only visual appraisal, without benefit of weight tapes, scales, or other formulae. The research results indicated that the 253 horsemen and veterinarians with an average of 15 years of experience substantially erred on the average when visually estimating equine weight. The average underestimation per horse was 166 lbs, or 16% of the average total body weight per horse. The results of this study indicate that underestimating the weight of a horse is a common error.
Traditional Equine Weight Estimation Methods
Due to the lack of availability or access to equine scales, and the general unreliability of visual estimation, the need for an accurate and effective equine weight estimation method or model exists. Over the past four decades, several weight estimation methods have been developed and tested, with varying degrees of accuracy.
These methods include girth formulae, girth weight tapes which are an embodiment of these formulae, two formulae involving use of girth and length as measured from the tuber ischium to olecranon (Ensminger-1973, and a traditional formula), one formula which utilizes umbilical girth and length as measured from the tuber ischium to the olecranon (Jones, et al-1989), and one formula utilizing girth and length as measured from the turber ischium to the point of the shoulder (Carroll/Huntington-1988).
EXISTING WEIGHT ESTIMATION FORMULAE:
Girth Weight Tapes (Commercial / Feed Companies)
Proprietary—Each Company Utilizes Different Girth Model Based On Individual Research And Data (Results Vary Significantly Between Tapes)
Weight (lbs.) = ‘Girth 1’ (ins.)2 x ‘Length 2’ (ins.)
Weight (kg.) = ‘Girth 1’ (ins.)2 x ‘Length 2’ (ins.) + 22.7
Carroll and Huntington (1988)9
Weight (kg.) = ‘Girth 1’ (cm)2 x ‘Length 1’ (cm)
Jones, et al (1989)10
Weight (kg.) = ‘Girth 2’ (cm)1.78 x ‘Length 2’ (cm).97
Hapgood, et al ‘New Method 1’ (2001- Patent Pending)
Weight (lbs.) = ‘Girth 1’ (ins.)1.64 x Height (ins.).95 x ‘Length 1’ (ins.).40
Hapgood, et al ‘New Method 2’ (2001- Patent Pending)
Weight (lbs.) = (‘Girth 1’ (ins.) x .53 + Height (ins.) x.34 + ‘Length 1’ x .13)2.99
Materials and Methods
Research conducted for development of the new methods detailed in this research paper was completed over a four-month period with the cooperation of the State Fair of Oklahoma. Participants in the State Fair of Oklahoma Horse Show allowed their horses to be weighed and measured. Data was gathered on 77 horses ranging in weight from 300 lbs. to 2,010 lbs., and in height from 10.5 hands to 18.25 hands. Data elements collected included girth (in.), height (in.), length (in.), visual weight estimate, actual weight (lbs.), age, breed and gender.
Three different commercial horse weight tapes were obtained and utilized in the research study. They included the Purina, Sure-Measure and Nutrena girth weight tapes. These tapes are normally used by wrapping the tape around the heart girth of the horse, and where the tape overlaps, the indicated weight is read from the tape. Each horse in the research sample database was measured with a regular measuring tape in inches, then weighed on a digital scale. Measurements of girth, height, and length were recorded. The girth weight tape results were extrapolated later by converting the girth measurements obtained in inches into the weight indications for each tape and the data was recorded.
The Girth1 (heart girth) measurement was obtained by taking a measurement in inches around the heart girth of the horse with the measuring tape snug against the horse following respiratory expiration. The tape was placed directly behind the elbows, the ends of the tape were overlapped, and the indicated heart girth measurement was recorded. The Girth2 (umbilical girth) measurement was obtained by taking a measurement in inches around the umbilical girth of the horse with the measuring tape snug against the horse following respiratory expiration. Once the tape was properly placed, the ends of the tape were overlapped, and the indicated umbilical girth measurement was recorded. The Height measurement was obtained by taking the distance in inches from the ground to the top of the withers. The Length1 measurement was obtained by taking the distance in inches from the point of the shoulder to the tuber ischium in a straight line. The Length2 measurement was obtained by taking the distance in inches from the olecranon to the tuber ischium in a straight line. A non-stretch fiberglass measuring tape was utilized during the research project.
Before each horse was weighed on the scale, the author made and recorded a visual estimate of the horse’s weight. The same researcher estimated the weight of every horse in the study. Each horse’s actual weight in pounds was recorded before returning them to their stalls.
Five new equine weight estimation models were constructed utilizing Multiple Regression Analysis (MRA). MRA is a statistical technique for estimating unknown data on the basis of known and available data. In equine weight estimation, the unknown data are horse weights. The known and available data are various measurements and characteristics of each individual horse. The objective of MRA, as applied to equine weight estimation is to model the relationship between horse characteristics / measurements and equine weight, so that the latter can be estimated from the former.
First, a correlation matrix showing the linear relationships between horse characteristics and actual weight was created. The Pearson Correlation Coefficients for each horse characteristic were then compared against actual weight and against each other to help determine what sort of relationships might exist between the independent variables which would potentially be utilized to predict the dependent variable of actual weight. As evidenced by the correlation matrix, girth was the most highly correlated with actual weight, followed by height, then length.
Using the variables girth, height and length, several new multiplicative models were developed by using the natural logarithm of actual weight as the dependent variable, and the natural logarithms of the various independent variables in a forward stepwise linear regression equation (criteria for probability of F to enter <= .050 and probability of F to remove >= .100). When converted, these produced the following formulae:
New Model 1:
Weight (lbs.) = ‘Girth 2’ (ins.).947 x Height (ins.)1.189 x ‘Length 2’ (ins.).930
New Model 2:
Weight (lbs.) = ‘Length 1’ (ins.)1.740 x Height (ins.)1.192 x ‘Girth 2’ (ins.).149
New Model 3:
Weight (lbs.) = ‘Girth 1’ (ins.)2.605 x ‘Length 2’ (ins.).122
New Model 4:
Weight (lbs.) = ‘Girth 2’ (ins.).270 x Height (ins.)3.237
New Model 5:
Weight (lbs.) = ‘Length 1’ (ins.)2.610 x ‘Girth 2’ (ins.).145
Discussion of Results
By introducing additional variables into estimation models, as done in Ensminger, Carroll/Huntington, Jones, and the models developed in phases II and III of this project with Heart Girth, Umbilical Girth, Height, Length from the tuber ischium to the point of the shoulder, and Length from the tuber ischium to the olecranon, more of these proportional differences are taken into account when predicting equine weight, improving accuracy levels.
Accuracy with girth weight tapes appeared to be lower for horses less than 500 lbs., and for horses in the 1,500 lb. range and higher, girth weight tapes could not be used at all because heart girth of the horses exceeded the length of the girth weight tapes (this is not a limitation for the various other formulae or the new methods detailed in this article).
The girth weight tape predicts only one weight for all the horses with the same girth measurement, even though height and length can fluctuate significantly among these horses. As a result of these other differences, there can be a significant range of actual weight of horses with the same girth measurement. The weight tape does not take these other differences into account, since the formula is based upon just the girth variable.
Although the new variables used in this year’s research project models produced good equine weight estimation results, it appears that Girth2 (umbilical girth) is not as effective as Girth1 (heart girth) in predicting equine weight. Additionally, it appears that the Length2 measurement (from the tuber ischium to the olecranon) is not as effective as Length1 (from the tuber ischium to the point of the shoulder) as a predictor variable.
The two models developed last year utilizing Girth1 (heart girth), Length 1 (from the tuber ischium to the point of the shoulder) and Height appear to be the best methods for estimating equine weight.
An equine weight estimation software program utilizing these formulae was created which allows a user to type a girth, height, and length measurement into the appropriate boxes, press <Calculate>, and have an accurate weight estimate appear on the screen. This same concept will be incorporated into an equine weight estimation calculator that can be utilized in the field, producing an immediate weight estimate without the necessity of a personal computer or laptop.
The software and calculator simplify the use of the mathematical formulae developed through this research project thus making it more practical for use by veterinarians, horse owners, trainers, and others involved in the care and management of the horse. The ultimate goal has been to create an accurate yet easy to use option for estimating equine weight.
Phase III Assistance, Fall 2001:
-Dr. Robert Curley, Business Statistics Professor, UCO
-Dr. Mark Bianchi, Practicing Equine Veterinarian
-Dr. Maria Luke, Practicing Equine Veterinarian
-State Fair of Oklahoma Board
-Pat Rush, State Fair of Oklahoma Horse Show Secretary
-Oklahoma City Mounted Police, with special thanks to Danny Booth
-Glenda Bales, Oklahoma Apaloosa Club President
-State Fair Horse Show Participants who volunteered their time and allowed their 77 horses to be weighed and measured.
-Kris and Joe Hapgood
1Giffin, James M., and Gore, Tom, (1998) Horse Owner’s Veterinary Handbook, 2nd ed., Howell Book House, N.Y., N.Y., pp 458-460.
2Hintz, Harold F. (1988) Weighing Horses, Equine Practice, 10, (8), 10-11.
3Thomas, Heather Smith, (2000), Storey’s Guide to Raising Horses, Storey Books, Pownal, VT, p 295.
4Milner, Jean, and Hewitt, D. (1969), Weight of Horses: Improved Estimates Based on Girth and Length, Can. Vet. Jour., 10 (12), 314-316.
5Ellis, J.M., and Hollands, T (1998), Accuracy of Different Methods of Estimating the Weight of Horses, Veterinary Record 143, 335-336.
6Asquith, R.L., Johnson, E.L., Kivipelto, J., and Depew, C. (1990), Erroneous Weight Estimation of Horses, Proceedings of the Annual Convention of the American Association of Equine Practitioners, 599-607.
7Lewis, Lon D. (1996), Feeding and Care of the Horse, 2nd ed., Lippincott, Williams & Wilkins, Media, Pa., pp 118-121, 266.
8Reavell, D.G., (1999), Measuring and Estimating the Weight of Horses With Tapes, Formulae, and by Visual Assessment, Equine Veterinary Education/AE, December 1999, 188-193.
9Carroll C.L. , and Huntington P.J. (1988), Body Condition Scoring and Weight Estimation of Horses, Equine Veterinary Journal 20 (1), 21-45.
10Jones, R.S., Lawrence, T.L.J., Veevers, A., Cleave, N. and Hall, J. (1989) Accuracy of prediction of the liveweight of horses and body measurements. Vet. Rec. 125, 549-553.