Check if nLTT statistic between posteriors differ using statistics

[richelbilderbeek]

For every simulation, I run at least two BEAST2 runs per alignment. It would be a good idea to check if these differ using statistics.

Because if they do not differ, I can just remove these replicates from the full pipeline!

[richelbilderbeek]

I've checked if the the distributions are normal, so I could do a t-test. I've found the nLTT statistic distributions are not always normal.

[richelbilderbeek]

I use a Mann-Whitney U test to see if nLTT statistics come from a same distribution. I expect this to be true between BEAST2 replicates, and I want to see this is true between alignments and species trees.

[richelbilderbeek]

It appears not all of them do so:

[richelbilderbeek]

This is some weird behavior:

[richelbilderbeek]

This is not a bug, it is a feature:

[richelbilderbeek]

So the answer is:

• sometimes yes
• sometimes no

Brilliant :confused:

[richelbilderbeek]

This may mean the ESSes are low?

Let's measure those, so up to #101 first!

[richelbilderbeek]

I close this Issue because:

• (1) the distributions may differ
• (2) #121 may cause them to differ

Confirming and fixing #121 may yield different results, which will be valid results.