Dear colleagues,

I have encoded a matrix of eight morphological traits related to inflorescence architecture in my beloved umbellifers. Among characters some are binary, while others are multinomial (from 3 up to 5 states possible). Polymorphism are also accepted (e.g. "0/1/3"). I wanted to check if some traits co-occur with each other and whether there is some phylogenetic signal in the data.

I decided to perform logistic PCA to answer these questions. Prior to the analysis I converted my data to dummy variables and I am afraid about eventual false correlation reuslting from different weights of analyzed characters (the more categories the more identical dummy variables encoding a single state from original matrix).

Do you have any idea how to weight characters or how to implement multinomial logit for the sake of logisitic PCA?