A good way would be to create as many variables as possible that map anything relevant, genes, upbringing, sexual and gender expression, etc., and then doing a PCA to reduce the defining vector to as few elements as possible.
I like how you think but I’m not sure if that alone will hold water. A variable can vary wildly even though it’s not very relevant to the property you’re interested in, and PCA would consider such a variable to be very significant. Perhaps a neural network could find a latent space. But ideally we want the components to have some intuitive meaning for humans.
A good way would be to create as many variables as possible that map anything relevant, genes, upbringing, sexual and gender expression, etc., and then doing a PCA to reduce the defining vector to as few elements as possible.
I like how you think but I’m not sure if that alone will hold water. A variable can vary wildly even though it’s not very relevant to the property you’re interested in, and PCA would consider such a variable to be very significant. Perhaps a neural network could find a latent space. But ideally we want the components to have some intuitive meaning for humans.