search

AlexiaJM / LEGIT

An R package for the Latent Environmental & Genetic InTeraction (LEGIT) model
GNU General Public License v3.0
12 stars 2 forks source link

Rescaling the weights for better interpretation #2

Closed AlexiaJM closed 7 years ago

AlexiaJM commented 7 years ago

When there are negative weights in a score, the maximum score is not 1 which complicates interpretation. What we could do is after convergence divide the weights of a score by (maximum possible value - minimal possible values) from the data or a user-given number. This should be an option called "rescale".

Ex: if weight_gene1=.40, weight_gene2=.40, weight_gene1gene2=-.20, then the lowest genetic score is 0 if gene1=0, gene2=0 and the highest score is .40. The rescaled gene weights would thus be weight_gene1_rescaled=1, weight_gene2_rescaled=1, weight_gene1gene2_rescaled=-.50.

AlexiaJM commented 7 years ago

Bad idea, in a lot of examples, it actually confuse interpretation. For example, in one environmental score, my weights become -2.5, 1.8, 1.1 instead of -.45, .34, .21.