I am working on a dataset with 65 individuals that belong to 6 species. I am trying to use a linear model that corrects for phylogenetic relatedness amongst individuals, however I am concerned that this is reducing the power of my models given that my effective sample size is small, due to having only 6 species. IS 6 species too few to incorporate phylogeny?
You could argue for not correcting for phylogeny with very few species (e.g.
If you work with cladistic morphology there is no problem with analysis even for small numbers of taxa. With respect to an outgroup, just identify uniquely shared apomorphic conditions within the ingroup taxa. 3 mammals and 3 fish is not a problem at all and would not necessarily lead to 'spurious' associations if analyzed with respect to an outgroup any more than addressing a larger number of taxa.
If you do an ordinary regression with phylogenetically clustered data (such as 3 mammals and 3 fish) you are likely to confound associations across and within clades resulting in 'spurious' associations. See Felsenstein 1985 or a textbook. But estimating phylogenetic signal in trait associations (to correct for relatedness) with such small number of species is likely to be very uncertain.
The only complication I would see with the example (3 mammals, 3 fish) is that 'fish' is not sufficiently defined. Any size of taxa has the potential of rendering '' associations. One might argue that if the number of taxa known to us, living or fossil, is so small relative to the number of species that have existed throughout, all results may be 'spurious'.
It is not important how fish are defined - I am referring to what Felsenstein called a "worst case" scenario with (any) two groups of moderately close relatives, https://www.jstor.org/stable/2461605.
- Badges
- Science topic
- Similar topics
- Biological Science
- Zoology
- Ornithology