Figure 1.
Estimates of a mutation rate on 1000 samples of size 50 of pairs mutant counts – final counts.
The horizontal line marks the true value. The first two boxplots correspond the traditional - and ML methods, which estimate the expected number of mutations from the sample of mutant counts, then divide by the final number of cells, supposed as known. On the next two boxplots, the estimates have been multiplied by the unbiasing factor (1). The last two boxplots use the full samples of pairs but no prior knowledge on final numbers. The best results are obtained by the maximum likelihood method (last boxplot). The
-method (label MLP0) performs nearly as well.
Table 1.
Table 2.
Table 3.
Parameters and notations for the mathematical model.
Figure 2.
Relative biases on estimates of a mutation rate.
Relative biases are plotted as a function of the coefficient of variation . The different curves correspond to
values of
from
to
. Red curves show biases of the
-method. For blue curves, the bias has been corrected by the unbiasing factor (1). The correction maintains the bias under acceptable values even for relatively large
and
.