INTERNATIONAL JOURNAL OF AGRICULTURE & BIOLOGY
ISSN Print: 1560–8530; ISSN Online: 1814–9596
10–583/EET/2011/13–3–419–422
http://www.fspublishers.org
Full Length Article
To cite this paper: Erat, S., 2011. Application of linear, quadratic and cubic regression models to predict body weight from different body measurements in
domestic cats. Int. J. Agric. Biol., 13: 419–422
Application of Linear, Quadratic and Cubic Regression Models
to Predict Body Weight from Different Body Measurements in
Domestic Cats
SERKAN ERAT
1
Department of Animal Breeding and Husbandry, Faculty of Veterinary Medicine, Kirikkale University, Yahsihan, Kirikkale,
Turkey
1
Corresponding author’s e-mail: serkanerat@yahoo.com
ABSTRACT
The aims of this study were to predict body weight (BW) from different body measurements and to determine the best
regression model for domestic cats. For this aims, a total of 48 adult Turkish cats (20 females & 8 males Turkish Angora; 13
females & 7 males Turkish Van) were used. In the study, wither height (WH), body length (BL) and head circumference (HC)
were assumed as independent variables, whereas body weight was used as dependent variable. Linear, quadratic and cubic
effects of the independent variables were included in the assumed model as Y= b
0
+ b
1
X + b
2
X
2
+ b
3
X
3
+ e. Where Y = body
weight; b
0
= the intercept; X = independent variables, (WH, BL, or HC); b
1
, b
2
and b
3
= regression coefficients and e = random
error. Conceptual predictive (Cp) and Akaike information criterion (AIC) were used to determine the most suitable model
among the assumed models. The model that has the smallest Cp and AIC values is the best model. The R
2
values from the
regression indicate the BL (R
2
= 0.50) to be moderately related to the BW. Neither the quadratic term nor the cubic term was
significant for all body traits, whereas the linear term was highly significant (p < 0.001) for all independent variables. Since the
maximum number of independent variables is three, there were seven possible different models. It can be concluded that cat
body weight was explained with the following model. (BW) = - 4.53 + 0.11 WH + 0.13 BL with p- values, <0.001, 0.0083,
and <0.001 for the intercept, b
1
and b
2
, respectively with R
2
= 0.57. © 2011 Friends Science Publishers
Key Words: Cat; Regression analysis; Body weight estimation
INTRODUCTION
In a domestic cat population, body weight is an
important trait that is used in evaluating body condition
(Erat & Arikan, 2010) and health status (Lund et al., 2005)
of cats, in computing dosages and in prescribing drugs.
Body weight and condition score are also often used for
assessing nutritional condition of dog and cat (Laflamme,
1997; Esfandiari & Youssefi, 2010).
A lot of techniques, which are simple or sophisticated
and expensive or inexpensive, are available to get
information on animal’s body traits. The easiest way to
assess an animal’s body mass is to weigh the animal.
However, under some situations scale may not be available
and prediction of body weight from body measurements
could be preferred practically (Latshaw & Bishop,
2001).
Multiple regression analysis has been used widely to
describe quantitative association between dependent (body
weight) and independent variables (hearth girth, body length
& wither height etc.) in animal studies (Cankaya, 2009).
Several studies on cattle, sheep and goat (Mohammed &
Amin, 1997; Wilson et al., 1997; Atta & El khidir, 2004;
Topal & Macit, 2004; Adeyinka & Mohammed, 2006;
Bagui & Valdez, 2007), dog and cat (Pendergrass et al.,
1983; Valdez & Recuenco, 2003), horse and donkey
(Pearson & Ouassat, 1996; Marante et al., 2007) and poultry
(Latshaw & Bishop, 2001; Grona et al., 2009) have been
conducted to predict body weight from body measurements.
The body weight estimation from using external body
measurements on domestic cats is scarce. Valdez and
Recuenco (2003) estimated the body weight of Philippine
domestic cats using some external body measurements.
Therefore, the present study was designed to predict body
weight of Turkish cats (Felis catus) from wither height,
body length and head circumference and to choose the most
appropriate regression model.
MATERIALS AND METHODS
Source of data: A total of 48 adult Turkish cats (20 females
& 8 males Turkish Angora; 13 females & 7 males Turkish
Van) were used (Fig. 1). All cats were solid white in color
and sexually intact. Averages of weight and age for these
cats were 3.39 kg and 2.71 years, respectively.
Measurements: Live body weight (BW), wither height