Path Analysis of the Relationships Between Various Body Measures and Live Weight of American Bronze Turkeys Under Three Different Lighting Programs

Transcription

1 TARIM BİLİMLERİ DERGİİ 005, 11 () ath Analyss of the Relatonshps Between Varous Body Measures and Lve Weght of Amercan Bronze Turkeys Under Three Dfferent Lghtng rograms Mehmet MENDEŞ 1 Al KARABAYIR 1 Akın ALA 1 Gelş Tarh: Abstract: In anmal breedng and genetcs, knowledge of causal effects may provde valuable nformaton on selecton. Dfferent contrbutons of explanatory varables to the trats ndcate that the breeder has to choose the varable wth the largest contrbuton n a selecton effort. In ths study, drect and ndrect effects of four explanatory varables (hank Length, Breast Length, Breast Depth and Breast Crcumference) nfluental on lve weght at week 30 n Amercan Bronze turkeys were nvestgated usng path analyss. Results of the analyses ndcated that the drect effects of shank length were the largest on lve weght. In general, ndrect effects of Breast Length and Breast Crcumference through hank Length were hghest among all ndrect effects. hank Length s the most nfluental varable and must be ncluded n the model n estmatng the lve weght at week 30. Key Words: Lghtng programs, causal effects, path analyss, Amercan bronze turkey Üç Farklı Aydınlatma rogramında Yetştrlen Amerkan Bronz Hndlernde Değşk Vücut Ölçüler le Canlı Ağırlık Arasındak İlşklern ath Analz le İncelenmes Öz: Hayvan ıslahı ve genetğnde, sebep-sonuç lşklernn blnmes seleksyon hakkında öneml puçları vereblr. ebep değşkenlernn üzernde durulan özellklere katkılarının farklı olması, söz konusu özellklere en fazla katkı yapan değşkenlern belrlenmesn gerektrr. Bu çalışmada, Amerkan Bronz Hndlernn tespt edlen dört özellğnn (nck uzunluğu, göğüs uzunluğu, göğüs dernlğ ve göğüs çevres), 30. hafta canlı ağırlık üzerne doğrudan ve dolaylı etkler z katsayısı metodu (ath analz) kullanılarak analz edlmştr. Analz sonuçları, canlı ağırlık üzerne en fazla doğrudan etkye sahp olan değşkenn nck uzunluğu değşkennn olduğunu göstermştr. Genel olarak, göğüs uzunluğu ve göğüs çevres değşkenlernn nck uzunluğu üzernden olan dolaylı etkler de daha yüksek bulunmuştur. Buradan hareketle, 30. hafta canlı ağırlığa en etkl değşkenn nck uzunluğu olduğu ve bunun modele dahl edlmesnn gerekl olduğu sonucuna varılmıştır. Anahtar Kelmeler: Aydınlatma programı, sebep-sonuç lşkler, path analz, Amerkan bronz hnd Introducton Man purpose of the anmal breedng practces s to mprove trats of economc value. Dependng on the speces, breeders try to mprove trats such as meat, mlk, egg and wool trats. These trats are closely related to explanatory varables such as age of dam, lghtng program, sex, breed, breast length etc., dependng on the speces (Gürbüz et al. 1999, Şahner and Görgülü 000). Investgaton of these relatons may provde essental nformaton to breeders n selecton, crossbreedng and n other practces. Dfferent explanatory varables may have dfferent contrbutons on the trat. Those wth a larger effect on the trat may be the most mportant to the breeder (İşc et al. 004). These varables may have ndrect effects on the trats through the other factors as well as drect effects. Lghtng s one of the most mportant envronmental factors that affect poultry performance as well other anmal speces. Dfferent lghtng programs affect the relatonshps among the varables (Hamlton and Kenne 1997, Yahav et al. 000, Mendes et al. 005a, Mendes et al. 005b). The causal effects needs to be nvestgated separately n dfferent lghtng programs to explan how the relatonshps among the varables change n dfferent lghtng programs. Determnng the varables most effectve or least effectve on the trats n queston may be benefcal n selecton programs. Explanatory varables may have drect or ndrect effects on the trats (L 1975, Düzgüneş et al. 1996, Keskn 1998). Usually, the drect effects are measured usng correlaton coeffcent. However, ndrect effects may confound the correlaton coeffcent. Ths may be because another unaccounted varable may be contrbutng to the correlaton coeffcent (Wrght 1934, Wrght 1960, Wrght 1968, ral and Kayaalp 1995). Usng smple correlaton coeffcent between trats and explanatory varables may not explan the relatonshps n all aspects and may be nadequate n nvestgatng the causal effects among the varables. In order to arrve at sold conclusons, correlaton coeffcent should be parttoned nto components and analyzed ust lke n analyss of varance (Alvn et al. 1975, Gürbüz et al. 1999). 1 Çanakkale Onsekz Mart Ünverstes Zraat Fak. Zootekn Bölümü-Çanakkale

2 MENDEŞ, M., A. KARABAYIR and A. ALA, ath analyss of the relatonshps between varous body measures and 185 lve weght of Amercan bronze turkeys under three dfferent lghtng programs In many felds, such as agrculture, socal scences and medcne, path analyss s wdely used to nvestgate drect and ndrect causal effects between trats and explanatory varables. In ths study, relatonshps between lve weght at 30 week and four body measures (hank Length, Breast Depth, Breast Length and Breast Crcumference) were nvestgated va path analyss. Maor purpose of ths study was to determne the explanatory varable that s most effectve on lve weght at slaughter and to order the varables n terms of ther contrbutons to the model. Materals and Methods The data were collected from 60 Amercan Bronze turkeys. The brds were 15-week old at the begnnng of the study. The study was carred out at the Üveck Research Center of the Çanakkale Onsekz Mart Unversty. The anmals were rased under ntensve condtons wth a lghtng program of 3L: 1D n the frst 55 days of the study. They were then allowed to go onto pasture. Three dfferent artfcal lghtng programs n addton to day lght were appled startng from 16 th weeks of age. Group I (control), group II and group III were treated wth lghtng programs 3L: 1D, 18L: 6D and 1L: 1D, respectvely. Each group had 10 male and 10 female turkeys. The starter and growth dets of the anmals ncluded 8% crude proten, kcal/me and % crude proten, kcal/me, respectvely. Wheat and water were offered ad lbtum to the turkeys when they returned back from the pasture under sem-ntensve condton. The brds were weghed before slaughter at 30 weeks of age. hank Length (1), Breast Length (), Breast Depth (3) and Breast Crcumference (4) were measured at 30 weeks of age also. ath analyss was used to analyze the data. ath analyss provdes a method to nvestgate drect and ndrect effects. It s an extenson of the ordnary multple regresson model. A regresson s done for each varable n the model as a dependent on explanatory varables (Anonymous, 004). The path coeffcent from an explanatory varable () to a trat (Y) can be shown as below: = Y. where; b Y : ath coeffcent, from to Y (=1,, 3, 4) Y. b : artal regresson coeffcent, : tandard devaton of, Y : tandard devaton of Y From here, ndrect effects of on Y through are calculated as follows: r ( Y. )( Y. ) where r s the correlaton coeffcent between th and th ndependent varables and Y. s the path coeffcent that ndcates the drect effects of th ndependent varable on the dependent varable. Coeffcent of determnatons can be parttoned nto ts components usng path analyss as follows: R = Y.1 + Y. + Y.3 + Y.4 + r 1 Y.1 Y. + r 13 Y.1 Y.3 + r 14 Y.1 Y.4 + r 3 Y. Y.3 + r 4 Y. Y.4 + r 34 Y.3 Y.4 where; Y. and r ( )( ) show İ Y. Y. drect and combned effects of explanatory varables (1,, 3 and 4) n contrbutng to the varaton of Y (Düzgüneş et al. 1987, Okut and Orhan 1993, Düzgüneş et al. 1996, Keskn 1998, Gürbüz et al. 1999). Varance Inflaton Factors (VIF) was calculated to measure whether there were any multcollnearty problems among the explanatory varables (Draper and mth 1998). Results Descrptve statstcs by lghtng programs and sex are gven n Table 1. The correlaton coeffcents between the trats and explanatory varables and those between the explanatory varables are gven n Table. The correlaton coeffcent between Lve Weght (Y) and hank Length (1) was large enough to be sgnfcant (< 0.05) for male turkeys rased n the 3L: 1D lghtng program and the Breast Crcumference (4) was hghly correlated wth Y (< 0.01) for female turkeys n the same lghtng program. In the 18L: 6D lghtng program, male lve weghts had the hghest correlaton wth 1 also. On the other hand, females had a conflctng result to the prevous lghtng program; 1 and varables had the hghest correlatons wth Y. In the 1L: 1D lghtng program, hank Length (1), Breast Length () and Breast Depth (3) varables had the hghest correlatons wth Y n males whle n females, none of the correlatons were large enough to be sgnfcant (>0.15). There were no multcollnearty problems among the explanatory varables, snce the VIF value was smaller than 10 n all cases (Draper and mth, 1998) (Table 3). Regresson equatons constructed for turkeys by lghtng programs are gven n Table 3. The coeffcents n these equatons are the path coeffcents (Wrght, 1960). ath coeffcents for turkeys rased under 3L: 1D lghtng program ndcated that the 1 varable had the largest effect on lve weght (Y) at slaughter for male turkeys. 3

3 186 TARIM BİLİMLERİ DERGİİ 005, Clt 11, ayı Table 1. Descrptve statstcs by lghtng programs and sex 3L:1D Lghtng program ex Male Female Male Female Trats Mn Max Mn Max ± ± Lve weght (Y) (gr) 764±35 545± hank Length (1) (cm) 16.45± ± Breast Length () (cm) 16.5± ± Breast Depth (3) (cm) 6.03± ± Breast Crcumference (4) (cm) 7.90± ± L:6D Lghtng program Lve weght (Y) (gr) 771±386 55± hank Length (1) (cm) 15.63± ± Breast Length () (cm) 16.15± ± Breast Depth (3) (cm) 5.58± ± Breast Crcumference (4) (cm) 7.30± ± L:1D Lghtng program Lve weght (Y) (gr) 698±8 5443± hank Length (1) (cm) 14.± ± Breast Length () (cm) 15.08± ± Breast Depth (3) (cm) 5.11± ± Breast Crcumference (4) (cm) 64.11± ± Table Correlaton coeffcents between trats by lghtng programs and sex rog 3L:1D Lghtng program ex Male Female Y Y * ** ** 0.75** ** * L:6D Lghtng program * ** * * L:1D Lghtng program ** ** 0.80** ** -0.81** -0.64* ** *< 0.05, ** < 0.01 Table 3. ath coeffcents, standard error, standardzed regresson model, ts R and varance nflaton factor (VIF) for 3L:1D, 18L:6D and 1L:1D lghtng programs 3L:1D Lghtng program Male Female ath Coef. E VIF R ath Coef. E VIF R ** * * ** Y = 0, , , ,175 4 Y = 0, ,17 + 0, , L:6D Lghtng program ** * * Y = 0, , , ,09 4 Y = 0, ,479-0, ,10 4 1L:1D Lghtng program ** * ** Y =0, ,38-0, , Y=0, ,598-0, ,186 4

4 MENDEŞ, M., A. KARABAYIR and A. ALA, ath analyss of the relatonshps between varous body measures and 187 lve weght of Amercan bronze turkeys under three dfferent lghtng programs had the least contrbuton to the model. 4 had a negatve effect on lve weght, but ths effect was not statstcally sgnfcant. On the other hand, n females under the same lghtng condtons, lve weght was affected largely by 4. ath coeffcents for turkeys rased under 18L: 6D lghtng program ndcated that the Lve weght of both male and female turkeys rased under ths lghtng program was affected mostly by the 1 varable. In studes related to lghtng programs, ths lghtng program was reported to be the most optmum program (Mendes et al., 005a). The path coeffcents for turkeys rased under 1L: 1D lghtng program ndcated that the 1 varable had the largest effect on lve weght at slaughter for male turkeys whle had the largest effect on lve weght at slaughter for female turkeys. Regresson coeffcents changed accordng to the lghtng programs and gender (Table 3). The 18L: 6D lghtng program seems to gve the most pronounced relatonshps between the trats and the explanatory varables. The coeffcents from tables and 3 were also used to calculate the drect and ndrect effects (Table 4). In Table 4, dagonal values are the path coeffcents that measure the drect effects of each ndependent varable on the explanatory varables. Values off dagonal measure the ndrect effects on the dependent varables. In male turkeys rased under 3L: 1D lghtng program, only the path coeffcent related to hank Length (1) was large (<0.01). Ths ndcated that one unt change n standard devaton n the 1 varable resulted n unt change n standard devaton n the Y. Indrect effects of 4 (0.61) and (0.53) on Y, through 1, was the hghest among all ndrect effects of on Y. Indrect effects of 3 were close to zero. The largest drect contrbuton to female lve weght was by 4 (0.944; <0.01), 1 (0.415) and 3 (0.417). 1 and 3 had the same level of drect contrbutons (<0.05). Ths mples that one unt change n standard devaton of Breast Crcumference (4) results n unt change n standard devaton of Lve Weght whle that n hank Length and Breast Depth results n 0.4 unt change n standard devaton of Lve Weght. Effects of (-0.17) were not large (=0.107). In ths lghtng condton, the only coeffcent that had large drect effects on both male and female turkeys was hank Length (1). 1 affected Lve Weght largely through 3 whle and 3 had the largest effect on Lve Weght through 4. 4 had the largest effect on Lve Weght through. In the 18L:6D lghtng program, only the drect effects of 1 (0.663) were sgnfcant n male turkeys whle drect effects of 1 (0.58) and (0.479) were sgnfcant n female turkeys (<0.05, Table 4). In male turkeys, one unt change n standard devaton of hank Length resulted n unt change n standard devaton of Lve Weght whle n females, that resulted n 0.58 changes n standard devaton of Lve Weght. In addton, one unt change n standard devaton of resulted n unt change n standard devaton of Lve Weght. In both genders, and 4 had larger effects through 1 compared wth 3. In the 1L: 1D lghtng program, drect effects of 1 (0.689) and (0.38) n male turkeys was sgnfcant, whle drect effects of (0.598) was sgnfcant n female turkeys (<0.01, <0.05). In male turkeys, one unt change n standard devaton of hank Length resulted n unt change n standard devaton of Lve Weght whle that of Breast Length resulted n 0.38 unt change n standard devaton of Lve Weght. In female turkeys, one unt change n standard devaton of Breast Length resulted n unt change n standard devaton of Lve Weght. Indrect effects of and 3 through 1 were smlar (0.551 and 0.558) n the male turkeys. Indrect effects of 1 and through 3 and 4 were close to zero. Drect and combned effects of the varables n contrbutng to the varaton of Y are gven n Table 5. In all lghtng systems, hank Length (1) had the hghest drect contrbuton to the varaton n Y. Combned effects of 1 and and combned effects of 1 and 4 on Y were the hghest among the varable pars. Regardless of the lghtng program, hank Length (1) had the hghest drect effects on lve weght n both sex, whle ths was more obvous n males. Ths ndcates that n estmatng the lve weght at 30 week, hank Length s the most nfluental varable and must be ncluded n the model. These fndngs are supported by Nestor et al., (001) and Mendes et al. (005a, 005b). Table 4. Drect and Indrect effects by lghtng program 3L:1D Varables Male Varables Female ** ** ** ** 18L:6D ** ** * L:1D ** * **

5 188 TARIM BİLİMLERİ DERGİİ 005, Clt 11, ayı Conclusons Table 5. Drect and combned effects of the explanatory varables n contrbutng to the varaton of lve weght (%) Lghtng rogram 3L:1D 18L:6D 1L:1D Drect Effects Male Female Male Female Male Female Y Y Y Y.4 Combned Effects 1 and and and and and and In all three of the lghtng programs, changes of hank Length n male turkeys resulted n sgnfcant changes n Lve Weght. Ths statement was true for female turkeys also, wth the excepton of 1L: 1D lghtng program. In addton, ndrect effects of Breast Length and Breast Crcumference through hank Length were hghest among all ndrect effects. hank Length can be the most nfluental varable and should be ncluded n the model n estmatng the lve weght at week 30. It should be kept n mnd that sample sze s effectve on the relatonshps between varables. Larger data sets gve more relable results. References Alvn, D. F. and R. M. Hauser The decomposton of effects n path analyss. Amercan ocologcal Revew 40: Anonymous 004. ath Analyss. date: (07 eptember 004) Draper, N. R. and H. mth Appled Regresson Analyss. John Wley and ons Inc., New York, UA. Düzgüneş, O., A. Elcn and N. Akman Hayvan Islahı, Ankara Unv. Zraat Fak. Yayınları: 1437, Ankara. Düzgüneş, O., T. Kesc, O. Kavuncu and F. Gürbüz Arastırma ve Deneme Metodları. Ankara Unv. Zraat Fak. Yayınları: 101, Ankara. Gürbüz, F., E. Başpınar,. Keskn, M. Mendeş and B. Tekndal ath analyss technque. 4. Natonal Bostatstcs Meetng, 3-4 eptember 1999, Ankara. Hamlton, R. M. G. and J. Kenne The effects of lghtng program. Ingredent partcle sze and feed form on the performance of broler turkey. Canadan Journal of Anmal cence, 77: Işç, Ö., C. Takma and Y. Akbas tudy on Factors Effectng 305-Day Mlk roducton of Holsten Fresan Usng ath Analyss. Natonal Anmal cence Meetng, 1-3 eptember 004, Isparta. Keskn, ath coeffcents and path analyss. Masters thess Ankara Unversty Graduate chool. (unpublshed). L, C.C ath Analyss-a prmer. The Boxwood ress, Calforna, UA. Mendeş, M., A. Karabayır, I. E. Ersoy and C. Ataşoğlu. 005 a. Three dfferent lghtng programs on lve weght change of bronze turkeys under sem-ntensve condtons. Archves of Anmal Breedng (n press). Mendeş, M., A. Karabayır, I. E. Ersoy and T. avaş. 005b. The relatonshp among pre-and post slaughter trats of Amercan Bronze Turkey. Archves of Anmal Breedng (n press). Nestor, K. E., J. W. Anderson and.g. Velleman Genetc varaton n pure lnes and crosses of large-boded turkey lnes.. Carcass trats and body shape. oult c. 80: Okut, H. and H. Orhan ath analyss and correlaton coeffcent. I. Natonal Econometr and tatstcs Meetng, 11-1 November 1993, Izmr. Şahner,. and Ö. Görgülü ath analyss and applcaton. Mustafa Kemal Unv. Zraat Fak. Dergs 5: ral, R. and T. Kayaalp ath analyss of the ndependent varables effectve on varous trats n honey bees of Trakya regon. Harran Ünv., Zraat Fak. Dergs, 1: Wrght, The method of path coeffcents. Annals of Mathematcal tatstcs 5: Wrght, ath coeffcents and path regresson. Alternatve on complementary concepts. Bometrcs 16: Wrght, Genetc and Bometrc Foundaton. The Unversty of Chcago ress. Volume 1. Chcago, UA. Yahav,.,. Hurwtz and I. Rozenbom The effect of lght ntensty on growth and development of turkey toms. Brtsh oultry cence 41: İletşm adres: Mehmet MENDEŞ Çanakkale Onsekz Mart Ünv. Zraat Fak. Zootekn Bölümü-Çanakkale

6