|Van Vleck, Lloyd|
Submitted to: Midwestern Section of the American Society of Animal Science
Publication Type: Abstract Only
Publication Acceptance Date: 11/1/2004
Publication Date: 3/1/2005
Citation: Van Vleck, L.D. 2005. Three designs for estimating variances due to competition genetic effects [abstract]. Midwestern Section of the American Society of Animal Science 83(2):40. Interpretive Summary: No interpretive summary is required.
Technical Abstract: Separation of (co)variance components due to direct and competition genetic effects and pen effects for models with imbedded competition effects is a challenge due to high degree of confounding of the effects in a pen. An earlier simulation with sires mated to five dams, each producing 10 full-sib (FS) progeny which were then assigned randomly to pens of size six, showed that those components would be partitioned. For this simulation, 60 sires were each mated to only one dam with dams unrelated. The 10 FS progeny were assigned to pens of size 10 in three ways: I) all to same pen, II) each progeny randomly to one of 60 pens, or III) 5 to pen 1 and 5 to pen 2 with another litter of FS furnishing the other 5 penmates for both pens 1 and 2. Number of replicates was 400 for each of 9 sets of parameters with statistical models for analysis including pens as random or as fixed effects. Of interest were 1) ability of designs to allow partitioning of the (co)variance components and 2) empirical standard deviations (SD) for estimated components. On average, all designs allowed estimation of covariance components with pens either random or fixed. The hypothesis was that the most difficulty (and greater variability of estimates) would be for Design I with each FS litter confounded with a pen effect. The result was the opposite: Design I empirical standard deviations of estimates were smallest for all components by multiples depending on true parameters and the component estimated with pens considered either fixed or random. Design II had somewhat smaller SD than Design III for estimates of direct genetic and residual variances. With pens considered random, SD for estimates of competition genetic variance, direct-competition covariance and pen variance were somewhat smaller for Design III than Design II. With pens fixed, those SD were smaller with Design II than Design III. Generally SD for estimates of competition variance and direct-competition covariances were smaller when pens were considered fixed rather than random. Design of pen assignment does not seem to be a simple problem when competition effects must be considered.