Prediction of Service Sire Fertility M.T. Kuhn 1 *, J.L. Hutchison 1, and J.S. Clay 2 1 Animal Improvement Programs Laboratory Agriculture Research Service, USDA, Beltsville, MD 2 Dairy Records Management Systems, Raleigh, NC.
Introduction Currently, two predictions of service sire (SSR) fertility are being done ERCR by DRMS ATA Bull Fertility Summary Request for AIPL to conduct a national evaluation for SSR fertility
Introduction ERCR Began in 1986 Published annually in Hoard’s Dairyman Currently includes data from - DRMS, Ag Source, MN DHIA - About 64% of US ATA Began in 2003 Somewhat less publicized than ERCR Primarily western herds
ERCR Trait Non-Return at 70 days (NR70) => Only first service used Linear animal model Factors HYM Milk Parity DIM SSR A cow PE cow SSR_age SSRxHerd Var(SSR) = I 2 ssr Var(A cow ) = A a 2
ATA Bull Fertility Summary Trait Conception (0, 1) on up to first 5 services Threshold model Factors SSR Cow Herd-Yr-Mo-Tech Herd-P-Milk Herd-DIM Herd-SSR_Age Var(SSR) = I 2 ssr Var(Cow) = I 2 cow
Overall Objective Develop national (phenotypic) evaluation for service sire fertility Data collection Trait definition Model development Appropriate edits for field data
Alternatives to be Investigated 1) ERCR 2) ATA Bull Fertility Summary 3) Alternative model(s) using all services Linear model Threshold model Linear model using predicted value for breedings with no pregnancy confirmation
Immediate Objectives 1) Compare models ERCR An alternative linear model (AIPL_LM) 2) Compare trait definitions for AIPL_LM All services First service only (NR70)
AIPL_LM Factors Included HYS A ssr ST-MO PE ssr Parity Stud Cow EFI ssr Milk F ssr DIM SSR_age Cow_Age Var(Cow) = I 2 cow Var(A ssr ) = A 2 a Var(PE ssr ) = I 2 PE Cow Effects Env’l Effects SSR Effects
Predictor for AIPL_LM CR =A ssr + PE ssr + Stud + ß 1 *EFI + ß 2 *F ssr + ß 3 *SSR_age
Data Breeding records from 4 DRPCs Years available: Date of breeding, service sire, service number recorded for each service on each cow-lactation No pregnancy check information available Last service on file, assumed successful 1997 ERCRs supplied by DRMS
Methods AIPL_LM predictions computed, for each bull, using records from 1995 to 1997 = CR AIPL-1 for first service only = CR AIPL-All for all services Each bull’s conception rate in 1998, adjusted for herd effects, was calculated = CR 98
Methods Criterion for comparison of predictors was correlation with CR 98 Cor(ERCR, CR 98 ) Cor(CR AIPL-1, CR 98 ) Cor(CR AIPL-All, CR 98 )
Results Min. No. of Mates ERCRCR AIPL-1 CR AIPL-ALL No. Bulls Correlation of Predictors with CR 98
Results Correlations Among Predictors Minimum No. of Mates ERCR, CR AIPL ERCR, CR AIPL-ALL CR AIPL-1, CR AIPL-ALL
Summary and Conclusions Use of multiple services improves accuracy. Model alterations can also improve accuracy. Agreement between alternative models and trait definitions increases with increasing number of mates. Modeling the SSR effect in separate components appears effective.
Summary and Conclusions Further modeling research Additional factors to include Modeling effects related to time Threshold vs. linear model Accounting for uncertainty of pregnancy status