jaxlogit._config_data.ConfigData

class jaxlogit._config_data.ConfigData(weights: ndarray | Series | Sequence[Any] | None = None, avail: ndarray | Series | Sequence[Any] | None = None, panels: ndarray | Series | Sequence[Any] | None = None, init_coeff: ndarray | Series | Sequence[Any] | None = None, maxiter: int = 2000, random_state: int | None = None, n_draws: int = 1000, halton: bool = True, halton_opts: dict | None = None, tol_opts: dict | None = None, num_hess: bool = False, set_vars: dict[str, float] | None = None, optim_method: str = 'L-BFGS-scipy', skip_std_errs: bool = False, include_correlations: bool = False, force_positive_chol_diag: bool = True, hessian_by_row: bool = True, finite_diff_hessian: bool = False, batch_size: int | None = None, setup_completed: bool = False)

Configurations for the fit and predict functions with default values.

Parameters

weightsarray-like, shape (n_samples,), optional

Sample weights in long format.

availarray-like, shape (n_samples*n_alts,), optional

Availability of alternatives for the choice situations. One when available or zero otherwise.

panelsarray-like, shape (n_samples*n_alts,), optional

Identifiers in long format to create panels in combination with ids.

init_coeffnumpy.ndarray, shape (n_variables,), optional

Initial coefficients for estimation.

maxiterint, default=2000

Maximum number of iterations.

random_stateint, optional

Random seed for numpy random generator.

n_drawsint, default=1000

Number of random draws to approximate the mixing distributions of the random coefficients.

haltonbool, default=True

Whether the estimation uses halton draws.

halton_optsdict, optional

Options for generation of halton draws. The dictionary accepts the following options (keys):

• shufflebool, default=False

  Whether the Halton draws should be shuffled.

• dropint, default=100

  Number of initial Halton draws to discard to minimize correlations between Halton sequences.

• primeslist

  List of primes to be used as base for generation of Halton sequences.

tol_optsdict, optional

Options for tolerance of optimization routine. The dictionary accepts the following options (keys):

• ftolfloat, default=1e-10

  Tolerance for objective function (log-likelihood).

• gtolfloat, default=1e-5

  Tolerance for gradient function.

num_hessbool, default=False

Whether numerical hessian should be used for estimation of standard errors.

set_varsdict, optional

Specified variable names (keys) of variables to be set to the given value (values).

optim_method{‘BFGS-scipy’, ‘L-BFGS-scipy’, ‘L-BFGS-jax’, ‘BFGS-jax’}, default=’L-BFGS-scipy’

Optimization method to use for model estimation.

skip_std_errsbool, default=False

Whether estimation of standard errors should be skipped.

include_correlationsbool, default=False

Whether correlations between variables should be included as explanatory variables.

force_positive_chol_diagbool, default=True

Whether to force positive diagonal elements in Cholesky decomposition.

hessian_by_rowbool, default=True

Whether to calculate the hessian row by row in a for loop to save memory at the expense of runtime.

finite_diff_hessianbool, default=False

Whether the hessian should be computed using finite difference. If True, this will stay within memory limits.

batch_sizeint, optional

Size of batches used to avoid GPU memory overflow.

setup_completed: bool, default=False

Whether the setup has already been completed

__init__(weights: ndarray | Series | Sequence[Any] | None = None, avail: ndarray | Series | Sequence[Any] | None = None, panels: ndarray | Series | Sequence[Any] | None = None, init_coeff: ndarray | Series | Sequence[Any] | None = None, maxiter: int = 2000, random_state: int | None = None, n_draws: int = 1000, halton: bool = True, halton_opts: dict | None = None, tol_opts: dict | None = None, num_hess: bool = False, set_vars: dict[str, float] | None = None, optim_method: str = 'L-BFGS-scipy', skip_std_errs: bool = False, include_correlations: bool = False, force_positive_chol_diag: bool = True, hessian_by_row: bool = True, finite_diff_hessian: bool = False, batch_size: int | None = None, setup_completed: bool = False) → None

Methods

__init__([weights, avail, panels, ...])

Attributes

avail
batch_size
finite_diff_hessian
force_positive_chol_diag
halton
halton_opts
hessian_by_row
include_correlations
init_coeff
maxiter
n_draws
num_hess
optim_method
panels
random_state
set_vars
setup_completed
skip_std_errs
tol_opts
weights