cv_kfold_bal: Cross validation for balancing weights K-fold cross...

In ebenmichael/ents: Maximum Entropy Synthetic Controls

Description

Cross validation for balancing weights K-fold cross validation, fit on K-1 folds, evaluate balance on Kth fold

Usage

cv_kfold_bal(outcomes, metadata, n_folds, hyperparams, link = c("logit",
  "linear", "pos-linear"), regularizer = c(NULL, "l1", "l2", "ridge", "linf"),
  normalized = TRUE, outcome_col = NULL, cols = list(unit = "unit", time =
  "time", outcome = "outcome", treated = "treated"), opts = list())

Arguments

outcomes	Tidy dataframe with the outcomes and meta data
metadata	Dataframe of metadata
hyperparams	Vector of hyper-parameter settings to consider
regularizer	Dual of balance criterion
normalized	Whether to normalize the weights
outcome_col	Column name which identifies outcomes, if NULL then assume only one outcome
cols	Column names corresponding to the units, time variable, outcome, and treated indicator
opts	Optimization options MAX_ITERS Maximum number of iterations to run EPS Error tolerance

Value

best hyper parameter

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