| Trait | GSM | α | r2 |  (SD) |  |
|---|---|---|---|---|---|
| Size | A | −0.5 | 0.073 | A 0.58 (0.14) | 11 |
| A + A2 | −0.5 | 0.093 | A 0.57 (0.15) | 11 | |
| A2 0.21 (0.51) | |||||
| Fertility | A | 1 | 0.012 | A 0.24 (0.24) | 0 |
| A + A2 | 1 | 0.029 | A 0.36 (0.21) | 3 | |
| A2 1.24 (0.87) | |||||
| (A + A2)rec | 1 | 0.064 | Aarm 0.44 (0.18) | 7 | |
| Acen 0.02 (0.07) |
Results are shown for additive (A) and additive-by-additive (A2) GSM, and for the most predictive model tested (if neither of the above), shown in bold. α is the scaling parameter from Speed et al. (2012), which determines the effect size expectation for markers as a function of allele frequency, where 0 is unweighted and smaller values assign greater weight to rare alleles (see Equations 1 and 2). Unconstrained REML estimates and SD are shown for components > 0 at convergence.
is improvement over the null model (likelihood ratio). is additive and additive-by-additive similarity defined at the level of recombination rate domains (tips, arms and centers, merged across chromosomes). See Equation S4 and associated discussion in File S22. GSM, genetic similarity matrices; LR, log likelihood ratios; REML, restricted/residual maximum likelihood.
