Closed dlaehnemann closed 8 months ago
Hi David,
I think you just hit the lower branch length limit.
A couple of things to try:
--blmin 1e-9
--model GTGTR+FO
Thanks, will try those.
Generally, do you expect the ascertainment bias correction to actually change branch length ratios, or will it mostly scale the whole tree?
I guess asc. bias should mostly affect tree scaling, but maybe have a look at this paper:
Turning off ascertainment bias did work, so this was the lower branch length limit, indeed. Many thanks!
This issue sounds similar to issue #131 , but this should be resolved in the version of RAxML-NG I am using (
v1.2.0
).This awkward result could very well be due to my data, but this might also be some bug. In either case, I just want to double-check here. Basically, I have a
CATG
file with the first header line being32 10638
. So32
taxa and10638
genomic sites with genotypes.As I do have missing data, and the wiki example of a
CATG
file doesn't have any missing data, I might have mis-encoded the missing data. I have usedN
for any taxon where there is no data to provide a most-likely genotype, as this sounded like the natural IUPAC code to use. And I put in a uniform likelihood of0.01
for each genotype in such cases, which should be fine from a computational/statistical point of view, right? But maybe this encoding is somehow not read correctly?Here's an example record:
Otherwise, here's the log of the run. Maybe I am also mis-configuring something, but I hope you could spot this from the log: