Fig. 6

Partitioning contributions to standard deviations for estimates of SNP effects. Residuals, group effects and a fixed effect are sequentially added to the basic SNP-only model (infection and recovery times assumed known). The corresponding increase in the SDs in the posterior distributions for SNP effects is investigated for (a) the susceptibility \({a}_{g}\), b the infectivity \({a}_{f}\), and c the recoverability \({a}_{r}\). For comparison, four different scenarios are investigated: a pure design (no dominance estimate) with respectively \({N}_{\mathrm{group}}=4\) (i.e. a single replicate of the basic design) and \({N}_{\mathrm{group}}=12\) (i.e. three replicates of the basic design) and a mixed design (no dominance estimate) with \({N}_{\mathrm{group}}=4\) (i.e. two replicates) and \({N}_{\mathrm{group}}=4\) (i.e. six replicates). In each case, ~ 1000 individuals were partitioned equally among the contact groups. The residuals were chosen to have the covariance matrix \({{\varvec{\Sigma}}}_{gg}={{\varvec{\Sigma}}}_{ff}={{\varvec{\Sigma}}}_{rr}=1\), \({{\varvec{\Sigma}}}_{gf}=0.3\), \({{\varvec{\Sigma}}}_{gr}=-0.4\), and \({{\varvec{\Sigma}}}_{fr}=-0.2\), the group effects had a SD of \({\upsigma }_{G}=0.2\), and the fixed effect (assumed to represent sex with gender randomly allocated) had a size \({b}_{g0}={b}_{f0}={b}_{r0}=0.2\). Results were found to be largely insensitive to these essentially arbitrary choices