Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

USE OF STANDARDIZED CANONICAL COEFFICIENT IN PREDICTING LIVE WEIGHT AT WEANING IN INDIGENOUS PIG Ogah D. M. ,V. E. Aya and M. O. Momoh Animal Science Dept College of agriculture pmb 33 lafia Nasarawa State , Nigeria Animal Breeding and Physiology Dept College of Animal Science University of Agriculture, Makurdi, Nigeria mosesdogah@yahoo.coms Abstract A canonical correlation analysis was used to examine the relationship between some traits at birth and at weaning in indigenous pigs, the traits measured, include live weight (lwt) , body length (bdl) , rump height (rh)and hearth girth (hg) were obtained from fifty two piglets at birth and at weaning at eight weeks period. The traits measured at birth period were one set of measurement (X -variable) and the same traits measured at weaning period were the second set of measurement (Y- variables) . Four canonical correlation were obtained (0.988, 0.973, 0.591 and 0.284) .Among the estimated coefficient only the first canonical variable was significant (p<0.05). Live weight and body length at birth had a positive effect on the live weight at weaning .These two traits can be used as early selection criteria for determining piglet weaning weight. Key words: Canonical correlation, body measurements, pig and factor loading Introduction Nigeria is estimated to have 4.4million pigs,( Shaib et al . 1997), about 78% of the population are found in the sub humid zone of northern and southern guinea savannah (Lufadeju et al 1995), most of the pigs reared in these areas are local breeds managed under extensive system. Improvement in production efficiency of pigs has been based on development and application of quantitative genetic principles of selection and mating system (Wheeler and Campion ,1993).The use of multivariate approach such as canonical correlation is currently been emphases . Canonical correlation analysis (CCA)was proposed by Hotelling 1935 in Thomson 1984 is a technique for describing the relationship between two variables sets by calculating linear combination that are maximally correlated, it also has the ability to deal with two variable set simultaneously and to produce both structural and spatial meaning .The application of CCA such as determination of the relationship between some traits measure pre and post slaughtering , milk and reproductive traits, body measurement at birth and six weeks of calves were discussed in other livestock (Alkandari and Joliffe 1997, Fouries et al., 2002 ,Cankaya et al., 2008). However the application of CCA in pigs in estimation of the relationship between some measurements at different period is not common The relationship between early performance or early morphological performance and subsequent morphological expression is a veritable area to exploit in meat producing animals as it is a function of its genetic constitution and determine the productivity of the animal.

Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

The aim of this study is to estimate the relationship between some traits which were obtain from local pigs at birth period and at eight weeks period and secondly to determine which of the traits can be used as early selection criteria for determining local pigs that have high body weight at weaning using CCA. Materials and Methods The animals used in this experiment were 52 piglets from ten sows raised extensively by native farmers in Lafia , Nasarawa State. The data was generated between April to September 2009. The traits measured include live weight , body length , rump height and heart girth, measured at birth period were included in the first variable set (X- variable set) and the same traits measured at weaning period were included in the other variable set (Y variable sets) Statistical method and data analysis Canonical correlation analysis (CCA) was used to investigate the relationship among the measurement at birth and at weaning period. These analysis was performed with SAS PRO> CANCORR (SAS,1999) .From CCA a linear association between predictor variables (measurement at birth ) and dependent variables (measurement at weaning week) were determine d .Canonical variables are linear combination of the original quantitative measurements that contain the highest possible multiple correlation with each group and that summarizes among class variation . The goal of CCA is to evaluate the relative contribution of each variable to the derived canonical function in order to explain nature of the relationship(s) consider the following two equations vm =amx1 +amx2+………..ampXp 1 w1 =bm1y1+ bm2y2 +………bmpyp
2

equation (1 ) and (2) gives the new variable vm and w1 which are a linear combination of the X (birth) and y (weaning) variables respectively. Let cm be the correlation between vm and wm.the objective of the canonical correlation is to estimate am1, am2 …….amp and bm1 , bm2 …..bmp such that cm is maximum. Equation 1 and 2 are the canonical equation vm and ws are the canonical variates and cm is the canonical correlation(Sharma, 1996) Results and Discussion Means and standard errors of the variables of the local pigs are presented in Table 1 . The mean birth weight and weaning weight obtained her are higher than what Ncube et al .(2003) obtained for local pig genotype in Zimbabwe (0.97) and (4.17) but the birth weight in this study is comparable to that of local pigs of Mexico (1.32) as reported by Mota et al 2002 with a superior weaning weight of (9.49). Bivariate correlations displaying the relationship among body measurements at birth and at weaning are given in Table 2. The highest correlation was predicted between rump height at birth and body length at weaning

Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

(0.814) while the lowest correlation was between body length at birth and rump height at weaning .There were positive relationship between the body measurements and live weight both at birth and at weaning. The relationship between the body measurement and weight at weaning were generally higher than that at birth. This similar finding was reported by Cankaya and Kayaalp ,(2007) Cankaya et al. (2008) who reported that live weight change is a frequently recorded variable in animal research and that body measurements such as body length , hearth girth and chest width are reported as important indicators of the live weight in animal growth traits. Table 3 showed the first canonical correlation coefficient was the only significant 0.988(p<0.05) among all estimated canonical correlation coefficient, from the likelihood ratio test. This finding is similar to those of (Tatar and Elicin, 2002 and Cankaya and Kayaalp, 2007).Standard canonical coefficient (canonical weight) were given for the first canonical variable (W1 and V1 ) in Table 4. Magnitude of the canonical coefficient signify their relative contribution to the correlated variate that is these coefficient indicate the effect of the live weight and body measurement of the pigs at birth and on the same traits at weaning . There fore the canonical variate (W1 V1) representing the optimal linear combinations of dependant and independent variables can be defined by using the standardized canonical coefficient given in Table 4. W1 = 0.72(lwt1)-0.30(bdl1)-0.70(rh1)+0.62(hg1) V1 =0.94(lwt)+0.21(bdl)-0.10(rh)-0.62(hg) .Accordingly the traits measured from the pigs at birth only body length and live weight have a positive effect on the live weight of the weaning pig, that means if the value of the body length and live weight increases the live weight of the pig at weaning will increase .Positive correlation was found to exist among the characters except for heart girth and rump height which is noticed at canonical loading and cross loading Table 5 and 6. Variables with larger loading contribute more to the multivariate relationship among the live weight and body measurements .At cross loading live weight and body length measured from the pigs at birth period contributed mostly to canonical variate W1 whereas at weaning live weight and heart girth provided the most contribute to canonical variate V1. This suggest that selection for live weight and body length at birth period will lead to increase in live weight at weaning , if increase in body weight is the aim of the selection. References Al-Kandari N. and I. T. Jolliffe (1997) Variable selection and interpretation in Canonical correlation analysis . Commu. California pp65 Cankaya S. and G.T. kayaalp (2007) Estimation of relationship between live weight and some body measurement s in German Farm x Hair Crossbred by canonical correlation analysis J.Anim. Prod. 48: 27-32 Cankaya S. , E. Yazgan, G.T .Kayaalp, Z. Gocmez and U. Serbester (2008) Canonical correlation analysis for estimation of relationship between some body measurement at birth and six month period in Holstein Friesian calves. Journ. Of Anim. And Vet Advances 7(8) 953-958.

Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

Fourie P.J., R.W.C. Naser, J.J. Oliver , and C. VanderWesthuizen (2002 ).Relationship between production performance, visual appraisal and body measurements of young Dorper Ram .South African J. Anim.Sc.32(4)256-262. Hotelling H. (1935). The most predictable criterion. J.Edu.Psychol.26:139-142 Lufadeju E. A , E. O. Otchere, N. B. Mijindadi, G. O. Oyebanji, E.O. Oyedipe and B. Masari (1995.) on-farm adaptive research methodologies for livestock Federal Agricultural coordinating Unit Mota D ., R. Ramirez -Necoechea, M. Alonso-Spilsbury, C .Garcia-Contreras (2003) Characterisation of the productive performance in family pig farmers located in Ayotzingo, state of Mexico http.www.cipav.org .co/lrrd/lrrd14/1/mota141 htm. Ncube M.,K. Dzama ,M. Chimonyo, A. Kanengoni and H. Hamudikuwanda (2003). Effect of boar genotype on reproductive performance of the local sows of Zimbabwe lrrd 15(2) 2003 SAS Institute. Inc. SAS users guide statistic sas institute. (1999). Cary NC Shaib N., A. Aliyu and J. S. Bakshi (1997). Nigerian National Agricultural Research strategy Plan 1996- 2010 Federal Ministry of Agriculture and Natural Resources Abuja Nigeria Sharma S. ( 1996). Applied multivariate Techniques. John wiley and sons inc .New York pp493 Tatar A. M. and A. Elicin (2002). The research on the relationship between body weight and measurement in suckling and fattening period by the method of canonical correlation in ile de France x Akkaraman (B1) male lamb j. Agric Sci, 8 (1) 67-72 Thompson B. (1984). Canonical correlation analysis. Uses and interpretation .sage publication. California pp69 Wheeler M. B. and D. R. Campion (1993). Animal production a long standing biotechnological success. America J . clinical Nutrition 58 (Suppl) 276-281

Table 1 : Descriptive Statistics measurements at birth and at weaning of indigenous pig. Parameter Live weight (lwt) kg Body length (bdl)cm Rump height (rh)cm At birth 1.34±0.05 23.30±0.48 20.35±0.35 At weaning 4.48±0.65 25.60±0.59 23.25±0.40 23.99±0.37

Hearth girth (hg) cm 20.80±0.20

Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

Table 2 ; Correlation matrix between the traits lwt bdl rh hg lwt1 bdl1 rh1 hg1

bd rh hg bwt1 bdl1 rh1 hg1

0.644 0.169 0.572 0.503 0.042 -0.331 0.726 0.212 0.620 0.501 0.096 0.007 0.532 0.093 0.479 0.814 0.038 0.623 0.158 -0.182 0.067 0.138 0.646 0.434 0.798 0.296 0.659 -0.141

lwt, bdl, rh and hg = traits measured at birth ,

lwt1, bdl1,rh1 and hg1 = measured at weaning

Table3; Summary result of canonical correlation analysis Pair of Can.Variate Can.l correlation W1v1 W2v2 W3v3 W4v4 0.988 0.973 0.591 0.284 eigen values 40.397 18.089 0.538 0.088 likelihood ratio 0.001 0.031 0.598 0.919 probability 0.037 0.103 0.687 0.537

Table 4; X lwt V1 0.94

Standardized canonical coefficient for canonical variables variable set bdl 0.21 rh -0.10 hg -0.62 W1 lwt1 0.72 Y variable set bdl1 -0.27 rh1 -0.70 hg1 0.62

Proceedings of 35th Annual conference of Nigerian Society of Animal Production 2010

Table 5; canonical loading of original variable with their canonical variables X variable set lwt V1 0.85 bdl 0.46 rh 0.10 hg -0.19 W1 Y variable set lwt1 0.54 bdl1 0.24 rh1 -0.35 hg1 0.71

Table 6; Cross loading of original variables with opposite canonical variables X variable set lwt bdl rh hg Y variable set lwt1 bdl1 rh1 hg1

W1

0.84

0.46

0.10

-0.19

V1

0.53

0.23

-0.35

0.70