kcde_point_predict_given_kernel_centers_and_weights: Make predictions from an estimated kcde model forward...

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

Description

Make predictions from an estimated kcde model forward prediction_horizon time steps from the end of predict_data, based on the kernel functions and bandwidths specified in the kcde_fit object. This function requires that the lagged and lead observation vectors have already been computed.

Usage

kcde_point_predict_given_kernel_centers_and_weights(kernel_centers_and_weights,
  kcde_fit)

Arguments

kcde_fit	is an object representing a fitted kcde model
train_lagged_obs	is a matrix (with column names) containing the lagged observation vector computed from the training data. Each row corresponds to a time point. Each column is a (lagged) variable.
train_lead_obs	is a vector with length = nrow(train_lagged_obs) with the value of the prediction target variable corresponding to each row in the train_lagged_obs matrix.
prediction_lagged_obs	is a matrix (with column names) containing the lagged observation vector computed from the prediction data. There is only one row, representing one time point. Each column is a (lagged) variable.
prediction_test_lead_obs	is a matrix (with column names) containing prediction target vectors computed from the prediction data. Each row represents one time point. Each column is a (leading) target variable.
normalize_weights	boolean, should the weights be normalized?

Value

a list with three components: log_weights: a vector of length = length(train_lagged_obs) with the log of weights assigned to each observation (up to a constant of proportionality if normalize_weights is FALSE) weights: a vector of length = length(train_lagged_obs) with the weights assigned to each observation (up to a constant of proportionality if normalize_weights is FALSE) centers: a copy of the train_lead_obs argument – kernel centers

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