Statistical determination of chick sex from pre-hatch egg measurements

This study was aimed at using weight and morphometric measures of hatching eggs to predict the sex of chicks they would produce a priori, by building a statistical discriminant function. The data for the study came from three different strains (Anak, Marshall and Ross) of broiler hens that have adapted well to the hot humid environment of South Western Nigeria. A total of 1826 hatching eggs of the three strains were incubated, out of which only 1002 were successfully hatched. Eggs for the study were appropriately tagged and hatched in individual hatching compartments to prevent crossing of the chicks at break-out, and were immediately tagged and sexed by anatomical inspection of vent by a professional. A Principal Component Analysis (PCA) indicated that the first two components accounted for 98% of the total variation, with high loadings from egg weight, length, width and shape index. These variables were used to build a discriminant function in predicting chick sex across the three strains. There were significant (P<0.01) differences in fertility and hatchability across the strains with fertility been highest in Marshall (86.67%) but least in Ross (52.77%) strain, whereas hatchability was highest in the Ross (82.22%) but least in Marshall (72.67%) strain. It was observed that the accuracy of the prediction method was consistently higher in male than female chicks across the three strains studied albeit with different precision across the three strains.


INTRODUCTION
The broiler birds evolved after several years of genetic improvement, and their relevance and success as a relatively cheaper and steady source of animal protein hinges on their shorter life span and higher meat yield.Much progress in the productivity indices of Broilers are now achieved through improvement of several environmental factors regarding their growth, health and maintenance.
Sex of the birds had been reported to be a significant factor (Rondelli et al., 2003) in the actualization of these improvements at various stages of the animal's life.It has been well established that productivity performances of broiler birds is highly impacted by the way they are reared, that is, either as straight-run (mixed sex) or separate-sex (Ayhan et al., 2004;Atasoy and Aksoy, 2005;Abanikannda et al., 2009).Although birds raised as separate sex averagely performed better than birds raised together without sexual delineation, due to the fact that there are minimal wastages resulting from consequences of peck order and sex difference.
However the additional cost of sexing chicks at hatch discourages most commercial operators from separating the birds according to sex.Sexing is done by colour, feather or anatomical inspection, with ease of determination decreasing in that order (Abanikannda et al., 2009).
In Nigeria where electricity is largely epileptic (USAID, 2006), the additional cost of incubating eggs incurred from diesel for generators to run incubators calls for a means by which cost of hatching eggs could be reduced.
Since male chicks attract less income when compared to female chicks at hatch, this study, therefore aims at seeking alternative non-invasive and less expensive means of predicting the sex of a chick using measurements and indices obtained from its egg.

MATERIALS AND METHODS
An initial total of 1826 viable hatching eggs were sourced from three strains of broiler parent stock comprising, Anak (606), Marshall (570) and Ross (650), from three different reputable farms in the humid tropical South West Nigeria.After an initial incubation period of 14 days, the eggs were candled and 1322 eggs cumulatively accounting for 72.4% of total eggs set, comprising Anak (485), Marshall (494) and Ross (343) were fertile.The Marshall strain was raised on litter and naturally pen-mated, the Anak strain was kept in cages but also mated naturally, whereas the Ross strain was kept in cages and artificially inseminated.
The procedures for egg identification, tagging, measurements and incubation are described in Abanikannda et al. (2009).The fertile eggs after candling at the 14 th day of incubation were placed in demarcated hatching chambers and covered with wire gauze to prevent the crossing of the chicks to other compartments at hatching.The sex of the chicks was determined by a trained professional, through anatomical inspection of the chick's vent on the day of hatch.Each chick sex was recorded against the egg's code that hatched in that compartment.In all, a total of 1002 chicks hatched and were weighed and sexed.
Measurements in this study were taken with digital weighing scale and vernier callipers and flex measuring tape.The indices were computed as prescribed by various authorities (Panda, 1996;Narushin, 2005).The six measurements taken include; egg weight, egg length, egg width, vertical circumference, horizontal circumference and egg weight at 18 th day of incubation, while the four indices computed include; shape index, egg surface area, egg volume and egg density.Due to the large number of measurements and indices (10), a principal component analysis was conducted and it was observed that the first two components accounted for about 98% of the total variation recorded in the study, with only egg weight, egg length, egg width and shape index having high loadings on the two components.
All statistical analyses were done using JMP (R) 2007 statistical software.Basic descriptive statistics of these three measures and index were computed for all eggs studied and the mean values for each strain were computed.
Discriminant Analysis seeks to find a way to predict a classification (X) variable (nominal or ordinal) based on known continuous responses (Y).It is used to classify observations into two or more groups if you have a sample with known groups and can also used to investigate how variables contribute to group separation (JMP, 2007).Discriminant analysis is a method of predicting some level of a one-way classification based on known values of the responses.The technique is based on how close a set of measurement variables are to the multivariate means of the levels being predicted.Measurement variables are specified as Y effects and the classification variable as a single X effect (JMP, 2007).
A discriminant function was computed to classify the eggs into respective sex (male or female) based on the measurements and index.Two thirds of the data were randomly selected across the three strains and used as training data for the discrimination, while the remaining one-third was used as test data for the discriminant analysis.
The discrimination Ratio, D, is computed: Where, M = estimated measurement variance; P = estimated part variance; E = estimated variance of measurement error, and T = estimated total variance.

Fertility
The mean fertility for all eggs (N=1826) determined at the 14 th day incubation was 72.40% (Table 1), with the highest (86.67%) recorded by Marshall, while Ross had the least of 52.77%.This obvious difference across the strains may be attributed to different breeding practices employed by the farms.The very low fertility of Ross may be due to the fact that eggs were collected only three days after the birds were artificially inseminated, which is rather earlier than prescribed waiting period.The Ross hens had been withdrawn from breeding programmes for some time and were only introduced to breeding for the purpose of this study.Fertility was greatly influenced (P<0.001) by the strain of bird, with each of the three strains being statistically (P<0.01)different from the other two strains.
Strain of hen was a significant (P<0.01)source of variation in hatchability, with the three strains exhibiting statistical (P<0.01)difference in hatchability.The Ross strain had the highest hatchability (82.22%), which implies that if adequate time lapse was allowed between insemination and egg collection, this strain seems to be the most promising of the three studied.

Sex ratio
The average sex proportion in the entire study was 53.7 and 46.3% respectively (Table 2) for male and female chicks.This proportion was fairly consistent across the three strains, and the sex ratio was independent (P>0.05)Descriptive statistics of all the ten measurements and indices are presented in Tables 2 and 3.It was observed that Ross had the highest values for all the four variables presented, while Anak consistently had the least.However, it is worthy of note that largest differences due to sex for all variables was recorded in Anak but it was not large enough to be statistically significant (P>0.05).

Pre-hatch sex determination of broiler chicks
The classification and its associated error rates are presented in Table 4.The accuracy of classification by sex in the entire study was 66.87%, while the accuracy for Anak, Marshall and Ross was 66.76, 63.23 and 69.15%, respectively.Amongst the three strains, the Marshall had the best accuracy for classifying male chicks (93.75%) and the worst accuracy (28.14%) for classifying female chicks.
Accuracy of prediction was higher for males compared to females within each of the strains and the combined data.This implies that male chicks are better predicted from egg measurements than female chicks.The consistently higher proportion for males in the original data may contribute to this enhanced accuracy for male.

Conclusions
The following conclusions and recommendations could be drawn from this study; Weight and morphometric measurements of hatching eggs promise to be good discriminating factors in predicting sex of chick prior to hatching and it provides a cheaper alternative to sexing chicks following hatch and can easily be determined by automating the procedure for measuring the eggs and computing indices.Reliability and accuracy of the method is dependent on breed and differences in the measured variables due to sex of the chicks.Sexing chicks at the hatchery would reduce wastages in feed consumption and feed conversion efficiency of broiler chicks during brooding and rearing.The method can be applied to other groups of birds where it is more profitable to have a particular sex over the other.This discriminating method should be further investigated to include more measures,

Table 1 .
Fertility and Hatchability of eggs by strain of hen.

Table 2 .
Mean ± standard errors of egg measurements by breed and sex.