- adds a function `choicemodels.tools.monte_carlo_choices()` for efficient simulation of choices for a list of scenarios that have differing probability distributions, but no capacity constraints on the alternatives
0.2.dev2
- adds a `probabilities()` method to the `MultinomialLogitResults` class, which uses the fitted model coefficients to generate predicted probabilities for a table of choice scenarios
- adds a required `model_experssion` parameter to the `MultinomialLogitResults` constructor
0.2.dev1
- improves the reliability of the native MNL estimator: (a) reduces the chance of a memory overflow when exponentiating utilities and (b) reports warnings from SciPy if the likelihood maximization algorithm may not have converged correctly
- adds substantial functionality to the `MergedChoiceTable` utility: sampling of alternatives with or without replacement, alternative-specific weights, interaction weights that apply to combinations of choosers and alternatives, automatic joining of interaction terms onto the merged table, non-sampling (all the alternatives available for each chooser), and estimation/simulation support for all combinations
- `LargeMultinomialLogit` class now optionally accepts a `MergedChoiceTable` as input
0.2.dev0
- adds additional information to the summary table for the native MNL estimator: number of observations, df of the model, df of the residuals, rho-squared, rho-bar-squared, BIC, AIC, p values, timestamp