getStackWeights: Generate Stack Weights

In joeornstein/SRP: Stacked Regression and Poststratification

Description

Main Stacking Procedure: Construct k folds and generate predictions from base models

Usage

getStackWeights(trainset, binaryDepVar = F, hlmFormula, lasso_vars,
  lasso_factors = NULL, forestFormula, mns_best = NULL, knnFormula,
  k_best, gbm_vars, gbm_factors = NULL, gbm_params, gbm_tune,
  nfolds = 5)

Arguments

trainset	The training dataset
binaryDepVar	A boolean denoting whether the outcome variable is binary or continuous. Used to specify tuning parameters for machine learning methods.
hlmFormula	A formula object for the hierarchical linear model
lasso_vars	The predictor variables to include in the LASSO model
lasso_factors	Which predictor variables need to be converted to factors before estimating the LASSO model
forestFormula	A formula object for the random forest model
mns_best	Optional min node size parameter for random forest model
knnFormula	A formula object for the KNN model
k_best	Optimal k parameter for KNN model
gbm_vars	The predictor variables to include in GBM model
gbm_factors	Which predictor variables need to be converted to factors before estimating the GBM
gbm_params	List of parameters for the GBM model
gbm_tune	xgb.tune object, containing the $best.iteration
nfolds	Number of cross-validation folds

Value

The optimal stacking weights (HLM, LASSO, KNN, Random Forest, GBM)

Note

trainset must include a variable called 'y', denoting the outcome

joeornstein/SRP documentation built on Oct. 15, 2020, 8:30 p.m.