Is “semi-additive binary coding” justified?

[HS6986]

Dear Professor Smith,

Brazeau et al. (2019) advocated that “additive binary coding” should perhaps be used instead of reductive coding for neomorphic characters, which is commonly employed. This was because, when using the Brazeau et al.'s algorithm, reductive coding for neomorphic characters will often artifactually lend support for clans that are united by nonderived conditions (typically absence), which clearly convey less information than derived conditions (typically presence), and additive binary coding will prevent this from happening.

However, as also recognized in Brazeau et al. (2019), there is a problem with binary additive coding as well; the loss of a principal character requires [the number of its subordinate characters whose states are derived + 1] steps, which seems to be unrealistic.

I have come up with a new approach “semi-additive binary coding” to this problem although I am not sure if it is a right one. In this coding scheme (or, rather, a matrix transformation scheme), states in neomorphic characters that would be coded as “-” in reductive coding are coded as “(-0)”. I think that this coding scheme may be defensible on the principle of averaging multiple contradictory matrices (Ashkenazy et al., 2019), and that it may avoid artifactual clustering of taxa with nonderived states. As such, it may recover more accurate trees than the two aforementioned approaches, although of course it would not be a very great way to handle the problem.

My question is: is my “semi-additive binary coding” justified? If it is justified, I will post some issues I have found in running a matrix with this approach in TreeSearch.

Thank you very much for your time and support.

[ms609]

It's an interesting idea, but my initial impression is that, in the absence of any leaves unambiguously labelled -, all cases of (-0) will be treated as if they were 0.

[HS6986]

The behavior you mentioned—in the absence of any leaves unambiguously labelled “-”, all cases of “(-0)” will be treated as if they were “0”—was indeed confirmed by Inapp and TreeSearch::PlotCharacter(), but it seems that using semi-additive coding in TreeSearch somehow recovers topologies closer to ones obtained with binary reductive coding than with binary additive coding.

Regarding the provision of actual cases, please allow me a little more time.

Thank you for your time and support.

[HS6986]

Dear Professor Smith,

Here is a folder of my re-analyses of a Juravel et al. (2023) matrix for testing purposes, showing that using semi-additive coding in TreeSearch somehow recovers topologies closer to ones obtained with binary reductive coding than ones with binary additive coding. This may suggest that using “semi-additive binary coding” was effective. Please see README.md in the folder for more details.

[ms609]

One other thing to note here is that due to a bug in Morphy, leaves labelled (-0) are not in fact treated as such (I think they end up being treated as 0, or possibly ? – I forget). Martin Brazeau was looking into this some time ago but I forgot how far he got – the issue may have lain with the nexus parser used to read the input data?

[HS6986]

Thank you for your reply and for providing the information.

I see—thanks for pointing that out. I think I'll avoid using (-0).

Thank you very much for your time and support.