Hello, I have used the Pagels 94’ extended version in corDISC with 3 binary traits and 1300 species. To be able to use it, I had to code two binary character out of a continuous. My continuous variable is substrate proportion data. So 100% is gymnosperm exclusivity, 0% is gymnosperm exclusivity. I made two binary character out of it, using different thresholds 10%, 20%, 30%. 40%, 50%. The resulting two character were XY, where X= angiosperm occurrence, Y= gymnosperm occurrence. Therefore a 11 would be occurence on both and e.g. 10 is occurrence primarily on a given substrate, where “primarily” is defined by the threshold. The third binary doesn’t matter in this case.
So my first question is: Is it valid to have a state that is coded by two binaries, like 11 that means “occurrence on both” because the meaning of the coding depends on both binaries and is therefore not “independent”? I have a bad feeling about coding that has a “meaning together”... The other question is about the results. I ran corDISC with all 5 different thresholds and the results are super different. Almost all of them range between 10 and 100 (100 was the upper bound). So the general question is: Is the method not reliable for different codings of a continuous? So is the methods just not the right one? Or should I argue for a threshold and take this as “the” result? Or take the mean of the resulting transition rates and give the range? But then statistics based on likelihoods are not applicable any more… And a last but general question: How prone is corDISC for unbiased sampling? So if one trait is much more abundant than the other. Hope you can help me. I know the second question is a more about “good” science... All the Best, Franz Franz Krah (M.S.) Personal Webpage: http://franzkrah.github.io <http://franzkrah.github.io/> University: http://www.biodiv.wzw.tum.de/index.php?id=18 <http://www.biodiv.wzw.tum.de/index.php?id=18> currently at Clark University [[alternative HTML version deleted]] _______________________________________________ R-sig-phylo mailing list - [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-phylo Searchable archive at http://www.mail-archive.com/[email protected]/
