fit_prog_seq2seq: Fit a seq2seq model with a feedforward net to fit E[Y(0)|X]

In ebenmichael/ents: Maximum Entropy Synthetic Controls

Description

Fit a seq2seq model with a feedforward net to fit E[Y(0)|X]

Usage

fit_prog_seq2seq(X, y, trt, layers = list(c(50, "relu"), c(5, "relu")),
  epochs = 500, patience = 5, val_split = 0.2, verbose = F)

Arguments

X	Matrix of covariates/lagged outcomes
y	Matrix of post-period outcomes
trt	Vector of treatment indicator
layers	List of (n_hidden_units, activation function) pairs to define layers
epochs	Number of epochs for training
patience	Number of epochs to wait before early stopping
val_split	Proportion of control units to use for validation
verbose	Whether to print training progress

Value

• y0hat Predicted outcome under control

• params Model parameters

ebenmichael/ents documentation built on May 31, 2019, 8:45 p.m.