NestedLogit.make_nest_probs

NestedLogit.make_nest_probs(U, lambdas, nest_indices, alphas_nest=None)

Calculate nest selection probabilities and conditional probabilities.

This computes: 1. P(nest) = exp(I_nest) / Σ exp(I_nest’) [nest selection probability] 2. P(y|nest) = exp(U_y/λ) / Σ_{j∈nest} exp(U_j/λ) [within-nest choice]

Parameters:

UTensorVariable

Systematic utility, shape (n_obs, n_alts)

lambdasTensorVariable

Nest-specific lambda parameters

nest_indicesdict

Mapping of nest names to alternative indices

Returns:

nest_probsdict

Nest selection probabilities

conditional_probsdict

Within-nest conditional probabilities