Performance comparison of the PIQMEE method and the CLASSIC method on full, unique and randomly subsampled data sets of various sizes. For each setting, 100 data sets were analyzed. All the summary statistics shown are only for the runs that reached ESS of 200 for each parameter. The total number of such runs out of 100 is shown in smaller, colored numbers on the x-axis below each figure. (A) The number of sequences analyzed by each method. For PIQMEE_all and CLASSIC_all, the method was considering all the sequences. For CLASSIC_unique and CLASSIC_random, the size of the data set was smaller. (B) The distribution of the CPU seconds elapsed until the runs reached ESS of 200 for all the parameters. (C) The clock rate used for inference, in units of substitutions/site/time unit. As all our data sets have the sequences sampled at one time point, we need to fix the clock rate to infer the tree height. The clock rate for all analyses was fixed to the values used for simulations. (D) The distribution of the normalized relative error of median posterior estimates of the tree heights (). The ideal case of error being 0 is indicated by the gray line. (E) and (F) The distribution of the effective reproductive number () and the death rate estimates. The top and the bottom of the bars represent the median value of the top and the bottom end of the 95% HPD intervals, respectively. The median of the medians of the posterior estimates is displayed as the darker-colored dot. The true value of the parameter is indicated by the gray horizontal line. In (A), (B), and (E), the y-axis has a log-scale. For all subfigures, the summary statistics for the PIQMEE method on the full data set (PIQMEE_all) are shown in red, for the CLASSIC method on the full data set (CLASSIC_all) in blue, for the CLASSIC method on the unique subset (CLASSIC_unique) in green, and for the CLASSIC method on the random subset (CLASSIC_random) in purple.
