Consider recursive filling-in of life habit traits
Closed this issue · 1 comments
The algorithm leaves blank any state that the chosen relatives lack information on, because they only choose the closest relative with some unknown states, and not all unfilled-in states. This is sound logic because it's propagating only those states where there is confidence, based on taxonomic relatives, of known states. However, other states are therefore left unfilled. For example, for any trilobite or any extinct crinoid subclade, traits that are not really known in the fossil record, such as reproductive mode (SEXL vs ASEX) are therefore left unpropagated.
Consider adding a recursive propagation algorithm, where once the first states are filled in using the closest pool of relatives, then a second (and subsequent, as needed) round of propagations is conducted among higher (and higher) relatives until all available states are filled in. This could require a relatively simple change to existing code. And any propagations will be conservative, because for the remaining states, we're using larger and larger taxonomic ranks, filling on only those that are very common (in the 'mode') or constant (in 'constant') among each higher taxon.
Just wrapped up in new version