Initialization

Suppose we would like to simulate a process of households deciding among 3 choices using the discrete choice model theory. The number of persons in each household should influence their decisions. We will model this behavior by multinomial logit using a system of utility functions:

$$\begin{array}{rcrr} v_1 & = &\beta_{01} & \\ v_2 & = & & \beta_{12}x \\ v_3 & = & \beta_{03} & + \beta_{13}x \end{array}$$

Here, $\beta_{01}$ and $\beta_{03}$ are alternative specific constants and $x$ is a household variable ``persons''.

The model is initialized by

>>> from opus_core.choice_model import ChoiceModel
>>> choicemodel = ChoiceModel(
                       choice_set=[1,2,3],
                       utilities = "opus_core.linear_utilities",
                       probabilities = "opus_core.mnl_probabilities",
                       choices = "opus_core.random_choices")

Thus, the model is composed by external implementations of model steps, such as computing utilities, computing probabilities and computing choices, specified as package.module_name. Those modules must contain a class of the same name and a method run() that performs the actual computation (see Section 24.4.3 and Sections 24.5.1-24.5.3 for more details).