Create a FrengressionSeq Model for Longitudinal Data — frengression

Constructs a frengression model for time-varying/longitudinal causal data. Maintains lists of T networks for each component, with sequential dependence across time steps.

Usage

frengression_seq(
  x_dim,
  y_dim,
  z_dim,
  T_steps,
  s_dim,
  num_layer = 3,
  hidden_dim = 100,
  noise_dim = 10,
  x_binary = FALSE,
  z_binary = FALSE,
  y_binary = FALSE,
  s_binary_dims = 0,
  s_in_predict = TRUE
)

Arguments

x_dim: Dimension of treatment at each time step.
y_dim: Dimension of outcome at each time step.
z_dim: Dimension of time-varying confounder at each time step.
T_steps: Number of time steps.
s_dim: Dimension of static baseline variables.
num_layer: Number of layers (default: 3).
hidden_dim: Hidden layer width (default: 100).
noise_dim: Noise dimension (default: 10).
x_binary: Logical; is X binary? (default: FALSE).
z_binary: Logical; is Z binary? (default: FALSE).
y_binary: Logical; is Y binary? (default: FALSE).
s_binary_dims: Number of binary baseline dimensions (default: 0).
s_in_predict: Include S in prediction models (default: TRUE).

Value

A frengression_seq object.