search

xavierdidelot / BactDating

Bayesian inference of ancestral dates on bacterial phylogenetic trees
https://xavierdidelot.github.io/BactDating
MIT License
80 stars 15 forks source link

Root-to-tip R2 value? #32

Closed nbawe closed 4 years ago

nbawe commented 4 years ago

I have question. Is it ok to use results when R2 value is small (0.12), but the p-value is significant?

I have also tested temporal signal significance using equal dates with same model and also random dates permutations with multiple runs. All the analysis indicates significant temporal signal.

Thank you

xavierdidelot commented 4 years ago

Neither a high values of R2 nor a significant p-value are really strong evidence, because the root-to-tip approach is very rough due to the fact that root-to-tip distances are not independent of one another.

The best test in my opinion is to make sure you get a larger clock rate in your analysis than in any analysis where the dates are permuted. Or even better, make sure that the 95% credible interval with correct dates does not overlap with any of the 95% credible intervals with permuted dates. For more information see:

Duchêne, S., Duchêne, D., Holmes, E. C., and Ho, S. Y. W. 2015. The performance of the date-randomization test in phylogenetic analyses of time-structured virus data. Molecular Biology and Evolution 32:1895–1906.

nbawe commented 4 years ago

Hello again!

What do you mean any of the CI95%? - Do you mean only check mu values CI95% overlapping and greater than permuted dates should be in original run?

Or should I check all values (likelihood, prior etc) CI95% overlapping?

Thank you!

xavierdidelot commented 4 years ago

I meant the 95% credible interval for just the substitution rate mu.

nbawe commented 4 years ago

Thank you for prompt answer and also for the publication you shared!