Description of the study area
This study was carried out in two districts of BGR from December 2017 to April 2018. The districts, Homosha and Bambasi (Fig. 1), were selected purposively because they are believed to be breeding tracts of Arab and Oromo goats, respectively. Homosha is located in semiarid agro-ecology. It extends from 6°44′ to 6°84′ north latitude and from 37°92′ to 38°6′ east longitude [17] with an average altitude of 1373 masl [18]. The temperature ranges from 20 to 30 °C and the mean annual rainfall is 700–1200 mm. The Arab goats predominate in this district. The second study area, Bambasi, is situated in the sub-humid agro-ecology and it is positioned at 9°45′ north latitude and 34°44′ east longitude with an elevation of 1668 masl [19]. The mean annual rainfall ranges from 900 to 1500 mm and the average temperature is 28 °C. The Oromo goats are predominant in Bambasi district.
Sampling techniques and sample size
Two districts (Homosha and Bambasi) were purposively selected for this study. Subsequently, four peasant associations (PAs)—the lowest administrative units in Ethiopia—were selected from each district. The number of sampled households in each PA was determined following the recommended formula [20]:
$$N = \,0.25/\left( {{\text{SE}}} \right)^{2} ,$$
where N = sample size and SE = standard error. To make the number of the sampled households from each PA proportional to the size of the corresponding PA, the probability proportional to size (PPS) sampling technique was employed. The PPS was based on the formula:
$$W\, = \,\left[ {A/B} \right] \times N$$
[21], where W = number of households to be calculated from each selected PA; A = number of households in each selected PA; B = total number of households in all eight selected PAs and N = the calculated sample size. Detailed information on the sampling technique and the number of sampled households is given in [22]. Overall, a total of 86 households from Homosha district and 163 households from Bambasi district were selected following the steps in [20] and [21]. Finally, three goats per household were sampled for qualitative records and quantitative measurements. This gives a total of 747 goats (258 Arab and 489 Oromo goats). The goats were classified into six age groups based on their dentition, i.e., kids (< 6 months), young (6–12 months), one pair of permanent incisors (1PPI) (1 year), 2PPI (2 years), 3PPI (3 years) and 4PPI (≥ 4 years) [23]. The kids and young goats were differentiated by asking the age of goats from owners while the goats in 1, 2, 3 and 4 and above years were differentiated by observing their dentition. From the total sample size, 629 (84.2%) were female goats. Pregnant does were excluded from measurement to avoid over estimation of body weight (BW) and linear body measurements (LBMs). In the current study, quantitative traits, except BW, are generally named as LBMs.
Data collection
Ten qualitative variables (coat color pattern, coat color type, head profile, horn presence, horn shape, horn orientation, ear orientation, wattle presence, ruff presence and hair type) were recorded by using the standard format adapted from [6] breed descriptor list.
Nine morphometric measurements were also taken from each goat early in the morning before they were released for grazing. The measurements were taken as described by [6]. They included body weight (BW), [the fasted live body weight (in kg)]; chest girth (CG), (circumference of the body (in cm) immediately behind the shoulder blades and perpendicular to the body axis); body length (BL), (horizontal distance (in cm) from the point of shoulder to the pin bone); wither height (WH), [vertical height (in cm) from the bottom of the front foot to the highest point of the shoulder]; rump height (RH), (vertical height from the bottom of the back foot to the highest point of the rump); chest width (CW), [width (in cm) of the chest between the briskets]; pelvic width (PW), [horizontal distance (in cm) between the extreme lateral points of the hook bone of the pelvis]; horn length (HL), [length of the horn (in cm) on its exterior side from its root at the poll to the tip]; and ear length (EL), [length (in cm) of the external ear from its root on the poll to the tip]. Body weight (kg) measurements were recorded using suspended spring balance in kg with a precision of 0.2 kg. The height measurements (cm) were taken using a graduated measuring stick while the length, width and circumference measurements (cm) were measured with plastic measuring tape. All measurements were taken after restraining and holding the goats in their natural position.
Statistical analyses
All the data collected during the study period were encoded and fed into MS-Excel (2010) and analyzed using R statistical software version 3.5.2, 2018 [24]. However, based on the nature of data, different R packages were used.
During the qualitative data analysis, ‘gmodels package’ [25] was used to calculate the frequency and percent of qualitative characteristics observed in the two goat populations. On the other hand, the quantitative data were analyzed using the ‘lsmeans package’ [26]. Tukey’s comparison test was used to compare the sub-factors that brought significant differences.
The statistical model used was:
$$Y_{ijk} = \,\mu \, + {\text{A}}_{i} \, + \,{\text{G}}_{j} + ({\text{A}} \times {\text{G}})_{ij} + e_{ijk} ,$$
where Yijk = the recorded k (body weight and linear body measurements) in the ith age and jth goat population; μ = overall mean; Ai = fixed effect of ith age (i = 1, 2, 3 and 4; 1 = 1PPI, 2 = 2PPI, 3 = 3PPI, and 4 = 4PPI); Gj = fixed effect of jth goat population (j = 1 and 2; 1 = Arab and 2 = Oromo); (A × G)ij = interaction effect of age with goat population; and eijk = effect of random residual error. Due to the fact that only a few male goats at older age classes (3PPI and 4PPI) were available in the study area, male animals were excluded from the model in the analysis of BW and LBMs.
Pearson’s correlation coefficient (r) values were computed to assess the relationship between BW and LBM using ‘dplyr package’ [27]. Live BW was regressed on LBMs using stepwise multiple linear regression analysis. The coefficient of determination (R2) was used to assess the accuracy of prediction equations between BW and LBMs. Furthermore, MSE (mean square of error) was calculated from each fitted regression equation. In the first step, all LBMs were entered together into the equation for each goat population. Then, a group of variables having the maximum R2 and minimum MSE were selected. In addition, Akaike’s information criteria (AIC) and the Bayesian information criteria (BIC) were considered. In the second step, the variables which were selected by maximum R2 and minimum MSE were entered together into the model to find the best fitted regression equation:
$$Y_{i} = \beta_{0} + \beta_{1} X_{1} + \,\beta_{2} X_{2} \, + \,\beta_{3} X_{3} + \beta_{4} X_{4} + \beta_{5} X_{5} + \beta_{6} X_{6} + e_{i} ,$$
where Yi = dependent variable (BW); β0 = intercept; X1,..., X6 = independent variables (CG, BL, WH, PW, HL and EL); β1,..., β6 = regression coefficients of the variables X1,..., X6; and ei = residual random error.