Material and Methods
The animal material consisted of 244 Thoroughbred horses of different ages that run in races in
Hippodromes organized by the Jockey Club of Turkey. The animals were randomly selected from horses
housed in Adana Yeşiloba Hippodrome, in Adana city, Turkey (37°0' N and 35°19' E), in 2013. Pedigree
(stallion, birth date, age, gender, and mother age) and racing (running year, hippodrome, race distance,
race duration, racetrack, and race type) information of the horses was obtained from the records of the
Board of High Stewards,
Ministry of Food Agriculture and Livestock of Turkey, and the Jockey Club of
Turkey.
All morphometric measurements (Table 1) were taken from the right side with the horse standing in a
normal position inside a fixed crush.
Of the 244 investigated animals, 159 horses that had at least three paternal half-sibs and three
official racing records were selected for the statistical analysis of race performance and effects of
morphometric measurements on race performance. Race performance (m/sec) was calculated based
on duration and distance for each race.
The least squares mixed models including fixed effects of factors was used, as shown below:
yijklm = μ + αi + βj
+ γk
+ δl
+ eijklm (for morphometric measurements),
in which yijklm = dependent variable, µ = overall mean, αi = fixed effect of stallion, βj
= the fixed effect of
gender, γk
= fixed effect of age, δl
= fixed effect of mother age, and eijklm = random error.
R. Bras. Zootec., 48:e20180030, 2019
Multivariate analysis of morphometry effect on race performance in Thoroughbred horses
Paksoy and Ünal 3
yijklmnoprs = μ + αi
+ βj + γk
+ δl
+ ηm + κn
+ σo
+ τp
+ ωr
+ eijklmnoprs (for race performance),
in which yijklmnoprs = dependent variable, µ = overall mean, αi
= fixed effect of stallion, βj = fixed effect of
gender, γk
= fixed effect of age, δl
= fixed effect of mother age, ηm = fixed effect of year, κn
= fixed effect of
hippodrome, σo
= fixed effect of race distance, τp
= fixed effect of racetrack, ωr
= fixed effect of race type,
and eijklmnoprs = random error.
The meanings of the factors used in the mix models given above are explained below:
Stallion: father of horses investigated.
Gender: male and female horses investigated.
Age: ages of horses investigated = 2, 3, 4, 5, and 6+ years old.
Mother age: mother ages of horses investigated = 2-5, 6, 7, 8, 9, 10, 11, 12, 13-15, and 16-19 years old.
Year: the years of horse racing; 2001-2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, and 2013.
Hippodrome: cities of hippodrome where horse races took place = Adana, Bursa, İstanbul, Ankara,
İzmir, Diyarbakır, Şanlıurfa, and Elazığ.
Race distance: between 800 and 2400 m, and run by the horses investigated from start to end points =
800-1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000, 2100, 2200, and 2400 m.
Racetrack: ground composition of racetrack = dirt and turf.
Race type: flat racing categories organized according to various characteristics. Maiden race: racing
joined by horses that have never won. Handicap race: racing in which different weights are loaded
on the horses, and these weights are determined by scores of official handicappers to equalize their
chances of winning; handicap 13 (horses with a score between 1-50); handicap 14 (horses with a
score between 1-65); handicap 15 (horses with a score between 1-75); handicap 16 (horses with a
score between 30 and 85); and handicap 17 (horses with a score between 40 and 100). Condition race:
racing where horses participated according to the total amount of lifetime earnings, and additional
weights are loaded on horses according to the total amount of lifetime earnings. The more the number
of condition race increases, the more total amount of lifetime earnings are needed to participate. Open
class: racing with high-performance horses carrying the same weight.
Statistical significances among the subgroups were determined by Tukey’s test at 5% significance
level. Pearson’s correlation coefficients were calculated among morphometric measurements.
Table 1 - Descriptions of morphometric measurements examined in the research
Morphometric measurement Definition
WH - withers height Vertical distance from the highest point of the withers to the ground
RH - rump height Vertical distance between the highest point of the sacrum and the ground
CG - chest girth Circumference around the chest from behind the scapula
CW - chest width Distance in the front side between the outer sides of the right and left articulatio humeri
FCW - front chest width Distance in the front side between the inner sides of the right and left articulatio humeri
CD - chest depth Vertical distance between the highest point of the withers and the sternum
NL - neck length Distance from the angulus mandibula to the scapula
SL - shoulder length Distance from the highest point of the withers to the caput humeri
LWR - length of withers to rump Straight distance between the end of the withers and the beginning of the rump
RL - rump length Distance from the tuber coxae to the tuber ischia
BL - body length Horizontal distance from the caput humeri to the tuber ischia
HW - head width Distance between the right and left processus lacrimalis rostralis
HL - head length Distance from the crista occipitalis to the os incisivum
CC - cannon circumference Circumference at the middle of the metacarpal bone
R. Bras. Zootec., 48:e20180030, 2019
Multivariate analysis of morphometry effect on race performance in Thoroughbred horses
Paksoy and Ünal 4
The Kolmogorov-Simirnov normality test for normal distribution fitness, the Kaiser-Meyer-Olkin test
for sample size adequacy, and Bartlett’s sphericity test were applied in the morphometric measurement
data.
Principal component analysis for morphometric measurements were performed, and then the
factor loadings were rotated by the Varimax method. The significance of the rotated factor loadings
was determined using the value of 0.45, which is the limit for n = 159. Factor eigenvalues greater than 1
were accepted (Alpar, 2013; Tabachnick and Fidell, 2014). For factor analysis, the basic factor analysis
equation was used, as follows:
Zp×1 = ƛp×m Fm×1 + ep×1,
in which Z = p×1 vector of variables, ƛ = p×m matrix of factor loadings, F = m×1 vector of factors, and
e = p×1 vector of error factors.
Multiple Linear Regression Analysis, using the model below, determined the importance of the effects
of the obtained factors on race performance (RP):
RP = a + b1
FS1
+ b2
FS2
+ b3
FS3
+ b4
FS4 + e,
in which a = regression constant value; FS = factor scores; b1
, b2
, b3
, and b4
= regression coefficients of
factor scores; and e = error term.
The t test was used for significance of the regression coefficients. The autocorrelation assumption was
determined by the Durbin-Watson test. Statistical procedures were carried out using SPSS software
(Statistical Package in Social Sciences for Windows, version 14.01).
Results
Least squares means (Table 2) were 169.34±0.52 cm for withers height, 187.66±1.12 cm for chest girth,
and 168.52±0.75 cm for body length.
Table 2 - Descriptive statistics and P-values for the morphometric measurements of the horse
Morphometric
measurement
Descriptive statistics
of all horses (n = 244)
P-value of the selected horses
for race performance (n = 159)
X±Sx (cm) Min (cm) Max (cm) CV (%) X±Sx (cm) Stallion Gender Age Mother age
WH 169.15±0.23 157 178 2.11 169.34±0.52 <0.001 0.021 0.355 0.312
RH 167.46±0.23 155 177 2.16 167.65±0.54 <0.001 0.169 0.386 0.267
CG 187.22±0.46 162 204 3.86 187.66±1.12 0.032 0.390 0.555 0.852
CW 42.33±0.16 34 47 5.86 42.24±0.38 <0.001 0.469 0.227 0.356
FCW 20.34±0.10 16 24 8.01 20.10±0.25 0.038 0.606 0.552 0.769
CD 84.88±0.27 70 93 5.03 84.88±0.64 <0.001 0.302 0.282 0.524
NL 57.83±0.29 44 68 7.80 58.43±0.70 <0.001 0.158 0.600 0.671
SL 72.10±0.20 63 79 4.29 72.38±0.44 <0.001 0.211 0.171 0.741
LWR 62.50±0.26 54 72 6.45 62.43±0.56 <0.001 0.599 0.542 0.731
RL 37.57±0.17 31 43 7.13 37.22±0.36 <0.001 0.185 0.530 0.393
BL 168.17±0.33 154 178 3.04 168.52±0.75 <0.001 0.459 0.646 0.303
HW 23.88±0.05 21 25 3.02 24.00±0.13 0.116 0.029 0.381 0.552
HL 52.95±0.13 48 59 3.44 52.93±0.28 0.192 0.102 0.146 0.338
CC 20.46±0.05 19 22 3.52 20.48±0.11 0.040 0.038 0.326 0.453
WH - withers height; RH - rump height; CG - chest girth; CW - chest width; FCW - front chest width; CD - chest depth; NL - neck length;
SL - shoulder length; LWR - length of withers to rump; RL - rump length; BL - body length; HW - head width; HL - head length; CC - cannon
circumference; CV - coefficient of variation.
Tukey’s test (P>0.05).
R. Bras. Zootec., 48:e20180030, 2019
Multivariate analysis of morphometry effect on race performance in Thoroughbred horses
Paksoy and Ünal 5
Age and mother age had no significant effects (P>0.05) on any of the investigated morphometric
measurements, while gender was pronounced on withers height, cannon circumference, and head
width (P<0.05). On the other hand, for stallion, a significant effect (P<0.05; P<0.001) occurred on all
morphometric measurements except head length and width.
Least squares means for race performance were 15.29±0.06 m/sec (Table 3). Race performance was
significantly influenced by stallion (P<0.01), gender (P<0.01), age (P<0.05), year (P<0.05), hippodrome
(P<0.001), race distance (P<0.001), racetrack (P<0.001), and race type (P<0.001), while mother age
had no marked effect on this trait.
In general, high and significant (P<0.05; P<0.01; P<0.001) correlation coefficients among morphometric
measurements in positive direction were calculated, varying from 0.117 to 0.679 (Table 4). The 14
morphometric measurements examined showed fitness to normal distribution by the KolmogorovSmirnov test (P>0.05). Sample size adequacy by the Kaiser-Meyer-Olkin test was 0.849. The significance
level of Bartlett’s sphericity test was P<0.001.
Four factors (FI, FII, FIII, and FIV) with eigenvalues >1 were attained as a result of the analysis with
PCA. The factors of FI (general size), FII (body thickness), FIII, and FIV explained 52.46, 15.50, 8.01, and
7.21% of the total variation of 83.19%, respectively. Ten loads in FI, five in FII, two in FIII, and two in FIV
were statistically significant (Table 5).
The regression coefficients obtained for the factors were not significant (P>0.05) (Table 6). The
Durbin-Watson test for autocorrelation yielded a value of 1.377. Correlation coefficients between race
performance and regression coefficients of FI, FII, FIII, and FIV were calculated as ‒0.012, ‒0.079,
0.022, and 0.050 (P>0.05), respectively.
Discussion
Size and morphometry are extremely important traits in nearly all horse breeds including
Thoroughbred, and numerous breed registries select horses on functional criteria and support the
breeding of horses with body types most convenient for those particular functions. Using many body
measurements from the head, neck, trunk, and limbs in a great number of horse breeds such as
Thoroughbred, Shire, and Friesian in USA showed that there was a high body size variation among
the horse breeds (Brooks et al., 2010).
Coefficients of morphometric measurements covered in this study were low and less than 10%,
which shows that the uniformity of morphological characteristics of the breed were rather high.
These findings agreed with the report by Brooks et al. (2010) in which the lowest variation for body
measurements among 65 horse breeds was observed in Thoroughbred horse breed.
In terms of morphometric measurements examined, males showed higher values than females.
However, the effect of gender on morphometric measurements was generally not very clear, given
that gender effects were only important for withers height, cannon circumference, and head width.
There was usually a slight increase in morphometric measurements as the animals grew older,
but none of these increases was statistically significant. This shows that Thoroughbred horses
generally complete their growth and development at the age of two, agreeing with the statement
that Thoroughbred horses are early-maturing. In fact, Thoroughbred horses start their racing life
one year earlier (two years old) than Arabian horses, the other breed used in flat racing in many
countries. A research by Anderson and McIlwraith (2004) found that the various body measurements
of Thoroughbred horses were similar at two and three years old
