Background

It has been mentioned that breed characterization is a basic step to approach the sustainable use of animal genetic resource (Lanari et al. 2003). In the State of México, México, the region with the largest goat production is located in the rural municipalities of Amatepec and Tejupilco, with about 59 % of the total state goat production (Rebollar-Rebollar et al. 2012), with a predominance of local goats (Montaldo et al. 2010). The main product of goat farming in this region is meat; consequently, the trait of major interest for genetic improvement is live weight (LW). However, for goat producers in these locations it is very difficult to measure LW because most producers lack scales in order to weigh their animals, and sale prices depend entirely on live weight. Slippers et al. (2000) consider that visual estimates are too subjective; however, Otoikhian et al. (2008) indicate that it is feasible to achieve acceptable accuracy after having repeatedly estimated LW in goats by visual assessment. Nsoso et al. (2004) mentioned that it is possible to estimate LW in rural areas from the combination of zoometric measures and regression models, as a quick practice, of minimum cost and high reliability.

In Mexico, there is very little information regarding genetic and phenotypic variability of local goat populations, and even more about the efficiency of zoometric measures to estimate these parameters (Hernández Zepeda et al. 2002; Vargas et al. 2007). Objectives of the study were: (a) to compare LW and zoometric measures of local goats in two locations, and (b) to fit the best regression equation for LW prediction using zoometric measures.

Methods

Animals and study area

The present study was conducted in 17 goat production units (PU) of Amatepec and Tejupilco, located southwest of the State of Mexico, whose main characteristics are shown in Table 1. This study was conducted from February 2009 to October 2010, using 318 3-year-old local does. The management system is semi-extensive, with day grazing and night confinement. There are no regular health control practices and objective of the PU are breeding, sale of kids, and cull animals at weaning.

Table 1 Number of goats, climatic characteristics, altitude above sea level, and predominant forage species in Amatepec and Tejupilco, state of Mexico
Full size table

Zoometric measures analyzed

LW (kg) of does was measured along with the following linear zoometric measures (cm): body length (BL), trunk length (TL), withers height (WH), hearth girth (HG), rump width (RW), rump length (RL), head length (HL), head width (HW), and ear length (EL). LW was measured using a hanging scale of 100 kg of capacity and 100 g of accuracy. Zoometric measures were performed with a measuring tape, a drill compass and a pediometer or a graduated measuring stick (Revidatti et al. 2007). Because pregnancy produces bias in some zoometric measures, especially in the regions of the chest and rump (Yakubu et al. 2011), only non-pregnant does were utilized in this study.

Statistical methods

To test whether differences existed in means of LW and zoometric measures between does of Amatepec and Tejupilco student’s “t” tests for comparison of means (Steel and Torrie 1980) were performed. Means and standard deviations were obtained and the Pearson phenotypic correlation matrix was estimated for LW and zoometric measures.

In order to reduce the number of zoometric predictor measures (ZM) to estimate LW an analysis of principal components (APC) was carried out. This analysis allows transforming the original ZM in a new set of orthogonal variables (uncorrelated) called principal components (PC), which are a linear combination of the original variables (Johnson and Wichern 2007). In addition, APC of zoometric measures has been used as a tool in the assessment of animal body size and shape in goats (Okpeku et al. 2011). Upon APC, 9 PC were determined, and their variances, relative percentages and cumulative variances explained by the PC were obtained. The calculations were performed with the procedure PRINCOMP of SAS (SAS 2001).

For each location two multiple linear regression models were utilized to predict doe LW. The first, using as predictor variables the original ZM selected with the procedure STEPWISE of SAS (SAS 2001) based on the significance of their parameters (P ≤ 0.05), and the second, using as predictor variables the transformed values (“scores”) of the PC, that were selected based on the percentage of the variance that the PC explained. In both models the variables were standardized with mean 0 and variance 1, for facilitating their comparison.

The models have the following structure:

$$ {\text{LW }} = {\rm{\upbeta }}_{ 0} + {\rm{\upbeta }}_{\text{i}} {\text{X}}_{\text{i}} + \cdots + {\rm{\upbeta }}_{\text{n}} {\text{X}}_{\text{n}} + {{\epsilon }}\;\;\;\;\;\;{\text{(Model }}\; 1) $$

where LW is live weight, β0 is the intercept of the regression equation, βi is the ith partial regression coefficient of the ith zoometric measure retained in the model (Xi), βn is the nth partial regression coefficient of the nth zoometric measure retained in the model (Xn), ε is the random error.

$$ {\text{LW}} = {\rm{ \upgamma }}_{0} + {\rm{ \upgamma }}_{\text{i}} {\text{PC}}_{\text{i}} + \cdots + {\rm{\upgamma }}_{\text{n}} {\text{PC}}_{\text{n}} + {\rm{\epsilon }}\;\;\;\;{\text{(Model 2)}} $$

where LW is live weight, γ0 is the intercept of the regression equation, γi is the ith partial regression coefficient of scores of the ith principal component (PCi), γn is the nth partial regression coefficient of scores of the nth principal component (PCn), ε is the random error.

The criterion to evaluate goodness of fit of models used was based on the R2 value (Kleinbaum and Kupper 1978).

Results

Descriptive statistics

Table 2 shows descriptive statistics for LW and zoometric measures of goats from the two locations studied. Amatepec does were heavier, had a longer body, trunk, rump, head, and ear, and also a higher hearth girth than Tejupilco does.

Table 2 Least-squares means of live weight and zoometric measures of local goats in Amatepec and Tejupilco, state of Mexico
Full size table

Phenotypic correlations

The matrix of phenotypic correlations between live weight and zoometric measures is shown in Table 3. The highest phenotypic correlations (P ≤ 0.01) were found between LW and HG (0.83), HG and TL (0.75), BL and TL (0.74), HG and WH (0.73), LW and BL (0.72), and LW and HG (0.71). Except for EL, LW was highly correlated (P ≤ 0.01) with all zoometric measures.

Table 3 Matrix of phenotypic correlations between live weight and zoometric measures of local goats in Amatepec (above main diagonal) and Tejupilco (below main diagonal), state of Mexico
Full size table

Principal components and explained variances

From LW and nine zoometric measures of Amatepec does 5 PC were extracted which accounted for 83.3 % of the total variance (Table 4). From the analysis of Tejupilco does only 4 PC’s were extracted, which accounted for 82.4 % of the total variance.

Table 4 Principal components (PC), variance, partially explained variance, and cumulatively explained variance from the principal components analysis carried out for zoometric measures of local goats in Amatepec and Tejupilco, state of México
Full size table

Selected live weight prediction equations

Under the situation of Model 1 from the multiple linear regression analysis, the best model for Amatepec does included TL, HG, RW, and HW, with a value of R2 = 0.82 (Table 5), while for Tejupilco does the best model included BL, HG, HW, and EL with a value of R2 = 0.78, HG and HW being common measures for both populations.

Table 5 Multiple linear regression (stepwise) of live weight on original (model 1) and orthogonal (model 2) variables of local goats in Amatepec and Tejupilco, state of Mexico
Full size table

Discussion

Live weight and zoometric measures

The longer ears of Amatepec does, in addition to their superiority in LW and other zoometric measures, suggest the possibility that they have a higher influence of the Nubian breed than Tejupilco does, a breed that has been classified as a dual-purpose goat breed (Merlos-Brito et al. 2008). Local does of the State of Mexico have slightly lower or larger measures in various zoometrical traits than local does in different regions of México (Hernández Zepeda et al. 2002; Vargas et al. 2007), Venezuela (Pariacote et al. 2004), Argentina (Revidatti et al. 2007), and Cuba (Chacón et al. 2011). These differences can be mainly attributed to climate-related factors which generally produce changes in the quantity and quality of vegetation in different agro-ecological zones, thus leading to different ecotypes (Dossa et al. 2007). In addition, they can be also an effect of natural or artificial selection.

Phenotypic correlations

High correlations were found between HG and BL with LW (P ≤ 0.01) in both populations (Table 3), which coincides with results obtained by Pariacote et al. (2004). This result suggests that producers who lack scales for weighing animals can estimate LW of their goats using either of those two zoometric measures; that is, they can use a tape rule instead of a weighing scale, a practice that is much easier to perform under field conditions.

Principal components and explained variances

Okpeku et al. (2011) compared West African Dwarf (WAD) and Red Sokoto (RS) goats in LW and four zoometric measures by performing an APC, and extracted 2 PC in each breed. Moreover, the 2 PC extracted in WAD goats explained 94.15 % of the total variance, while the 2 PC extracted in RS goats explained 91.25 % of the total variance. Pariacote et al. (2004) performed an APC with 10 zoometric measures in local goats and extracted 10 PC, which explained 81 % of the total variation. López-Carlos et al. (2010) indicated that the first PC represent general body size, while the second PC represent body shape or conformation.

Selected live weight prediction equations

Regarding HG, Leng et al. (2010) recommended the use of this trait as a reliable measure to predict LW in goats under field conditions, due to the fact that muscle, some fat, and bone structure contribute to its formation. Bello and Adama (2012) also found that HG was the best zoometric measure to predict LW in goats. Similar to the results of Tejupilco, Ribeiro et al. (2004) found that the best model to predict LW in local goats and their crosses included BL and HG.

Under the situation of Model 2 of the multiple regression analysis, the best model for Amatepec goats included PC1, PC2, PC7, and PC8, with a value of R2 = 0.76 (Table 5), while for Tejupilco goats the best model included PC1, PC5 and PC9, with a similar R2 value.

From a global analysis, considering situations of Models 1 and 2 from the multiple linear regression analysis, we conclude that the model to be used for Amatepec goats is that one that includes original variables as predictors of LW, due to their higher value of R2 (0.82), compared to the model that includes orthogonal variables as predictors of LW, which had a lower value of R2 (0.74). For Tejupilco goats the value of R2 was similar (0.76) under the situations of Models 1 and 2 from the multiple regression analysis, thus to predict LW of goats in this population either model can be used.

Conclusion and implications

There are significant (P ≤ 0.01) differences between local does of Amatepec and Tejupilco in live weight, body length, trunk length, hearth girth, rump length and width, head length, and ear length. Trunk length, hearth girth, rump width and head width are zoometric reliable measures for predicting live weight of Amatepec local does, while for Tejupilco does body length, hearth girth, head width, and ear length are the most informative measures. Since hearth girth and head width are measures common to both populations, producers who lack scales for weighing animals in those locations are therefore recommended to utilize those measures to estimate live weight of goats with a good reliability. In addition, they will find that the use of a tape rule or graduated measuring stick signifies the saving of a considerable time and labor, especially when they are ready to sell goats based on live weight.