genome.models.logit

Model constructors for logit type models

The following are some of the logit type models already implemented: - multinomial logit

Todo

• latent class logit
• ordered logit (P-logit, B-logit)
• nested logit
• mixed logit

class genome.models.logit.MultinomialLogit(input, n_vars, n_choices, beta=None, asc=None, ref=None)

Bases: genome.models.base.BaseModel

Initializes a standard Multinomial Logit (MNL) model

Parameters:

• input – A symbolic Tensor input
• n_vars (int) – Number of input variables
• n_choices (int) – Number of choice alternatives
• beta – \(\beta\) parameters - A TensorSharedVariable (optional)
• asc – Alternative Specific Constants - A TensorSharedVariable (optional)

errors(y)

see: errors()

negative_log_likelihood(y)

see: negative_log_likelihood()