create_kcde_control: Assemble a list of kcde_control parameters for the kcde...

In reichlab/kcde: Kernel conditional density estimation with flexible kernel specifications

Description

Assemble a list of kcde_control parameters for the kcde function with user-specified values.

Usage

create_kcde_control(X_names, y_names, time_name, prediction_horizons,
  kernel_components, filter_control, crossval_buffer, loss_fn,
  loss_fn_prediction_args, loss_args, variable_selection_method = "stepwise",
  par_packages = NULL, par_cores = 1L, na.action = "na.omit")

Arguments

X_names	a character vector of length >= 1 containing names of variables in the data data frame to use in forming the lagged observation process used for calculating weights
y_names	a character vector of length 1 containing the name of the variable in the data data frame to use as the target for prediction
prediction_horizons	integer vector: the number of time steps between the last observation and the time at which we make a prediction
kernel_components	a list with one component for each component of the kernel function. Each component is a list with the following entries: - vars_and_lags: a data frame with two columns: "var_name" and "lag_value". Each row specifies a combination of variable and lag that is included in this component of the kernel function. - kernel_fn: a function to evaluate the kernel - theta_fixed: a named list of parameters to kernel_fn whose values are held fixed (i.e., not estimated) - theta_est: a named list of parameters to kernel_fn whose values are to be estimated - initialize_theta_fn: a function to initialize both theta_fixed and theta_est - initialize_theta_args: a named list of arguments to initialize_theta_fn - vectorize_theta_est: a function that converts theta_est into an ordered vector on a scale suitable for passing as the first argument to optim. Required to return a list with three components: (1) theta_est - vector of parameters to be estimated (2) lb - vector of lower bounds to theta_est (3) ub - vector of upper bounds to theta_est - update_theta_from_vectorized_theta_est: a function that updates theta_est (in list form) from theta_est (in vector form).
crossval_buffer	during cross-validation, the number of indices before the time at which we are making a prediction to drop from the "training examples".
loss_args	a named list giving arguments to the loss function
variable_selection_method	a character string specifying the method used to determine variable inclusion
par_packages	a character vector containing names of packages that need to be loaded in instances of R when computations are performed in parallel.
na.action	a character string specifying how NA values should be handled.
loss_fn_name	a string giving the name of the function use to compute loss from predictions

Value

the (at this point, unvalidated) list of kcde_control parameters

reichlab/kcde documentation built on May 27, 2019, 4:53 a.m.