For one set of parameter settings, the data set contains intensity values for all detected retention time/mass pairs of the 20 samples.
The experimental results are obtained by using the set of parameter settings summarized in Table 1.
In Table 3, we report the simulation power at the 10−3, 10−4, and 10−5 significance levels for each set of parameter settings.
We also report the asymptotic power at each of the significance levels, determined by computing the non-centrality parameter (equation (A1)) for each set of parameter settings.
In Table 1, we report the empirical type I error rates at the 0.975, 0.10, 0.05, 0.025, and 0.01 significance levels for each set of parameter settings.
As in Table 1, in Table 2, we report the empirical type I error rates at the 0.10, 0.05, and 0.01 significance levels for each set of parameter settings.
This benchmark was repeated for numerous sets of parameter settings and also used for parameter optimization (benchmark 2).
We present the number of contigs, average contig size (in parentheses), and the number of reads assembled into contigs for each different set of parameter settings.
The whole process is repeated for all subjects, and the mean values of spiked compounds and false positives retrieved of all subjects are taken and collected for each set of parameter settings.
The same versions and the same sets of parameter settings for BWA-mem, Picard, and Scalpel, were also used in the rest of the study, including the analysis of WGS/WES data, standard WGS, and PCR-free data.
Therefore, it can be concluded that our proposed algorithm is not computationally expensive as less than 1% of CPU usage is reported for all ranges of parameter settings for this given data set.
