Differences in expected log pointwise predictive density (ELPD) from the best-fitting model, estimated by the leave-one-out (LOO) method for all models and all antigenic distance metrics. We fit the models separately for each antigenic distance metric, so comparisons are shown separately. The ΔELPD is the difference in ELPD between the LMM and the GAMM, so a positive number indicates the LMM performed better than the GAMM, and a larger number means the LMM outperforms the GAMM more. We show the ΔELPD ± its standard error, along with the ratio of the estimate to its standard error.

Differences in expected log pointwise predictive density (ELPD) from the best-fitting model, estimated by the leave-one-out (LOO) method for all models and all antigenic distance metrics. We fit the models separately for each antigenic distance metric, so comparisons are shown separately. The ΔELPD is the difference in ELPD between the LMM and the GAMM, so a positive number indicates the LMM performed better than the GAMM, and a larger number means the LMM outperforms the GAMM more. We show the ΔELPD ± its standard error, along with the ratio of the estimate to its standard error.