Notes on the MLOGD 'Likelihood ratio' plot:

This is a plot of the MLOGD scores for each reference sequence - non-reference sequence pair, as a function of pairwise sequence divergence (red circles).

The MLOGD score is the sum of the null versus alternate model log likelihood ratios over the whole pairwise sequence comparison, divided by the number of nucleotides used (i.e. column 6 of the statistics summary table). Positive values favour the alternate model while negative values favour the null model.

The blue circle (if shown) represents the sum over the phylogenetic tree (details). Sometimes the blue circle will be outside the plot range (e.g. if the summed divergence over the tree is > 0.8 mutations per column). However, the coordinates of the blue circle are also given below the plot.

Note that all these scores are per nucleotide likelihood scores. A very short pairwise alignment can, by chance, have a few particular mutations that lead to an unusually large per nucleotide likelihood score. For a longer pairwise alignment, or an alignment of many sequences, the per nucleotide likelihood score is likely to be more accurate. Thus it is the total log likelihood score, rather than the per nucleotide likelihood score, that should, in the end, be used to put a value on how strongly the input sequence data discriminates between the null and alternate models.