Suppose that for n randomly chosen traits, the correlation cor DGV, EBV) t has been estimated.
In order to obtain a genome-wide threshold we determined the 95-percentile permutation threshold among 500 randomly chosen traits and estimated the 95% upper-bound from the traits distribution.
While multitrait analysis typically identifies QTL associated with a group of traits without readily disclosing which traits are associated with the QTL, variable selection chooses traits that are statistically most significant and thus most likely involved in a biological process (section 9 in File S1).
While we did not measure community assembly processes directly, these chosen traits are important for various assembly processes and population dynamics such as competition, facilitation, productivity, and stress tolerance and longevity (Grime 1977; Westoby 1998; Cornelissen et al. 2003; Schöb et al. 2012; Michalet et al. 2014).
Many other cat breeds are the result of breeders deliberately choosing traits in their breeding cats to shape the head, body, and tail types, colors and patterns, sizes, and coat lengths of the future litters.
To correct for multiple-traits testing, we estimated the minimum P-value from each randomly chosen trait and estimated the corresponding q-value [ 43].
Thus, for r t Val = 1, the expected correlation between EBV and DGV equals the expected accuracy of DGV for a randomly chosen trait with r t Train specified.
We predicted the expected correlation between EBV and DGV of a randomly chosen trait by assuming the following linear regression model (2) cor DGV, EBV t = a 0 + a 1 r t Val + a 2 r t Train + e t where the intercept a0 and the regression coefficients a1 and a2 are fixed effects, and the errors e t are normally distributed.
In the 'Search by Trait page', users can choose individual traits or groups of traits within specified trait thresholds.
