init_stacker: Initialize stacker

In dahcase/mbgstacking: Stacked generalization ensemble model for model-based geostatistics

Description

Creates a stacker governor containing the various information and data to run stacked generalization in IHME's MBG framework

Usage

init_stacker(..., inlist = T, data, indicator, indicator_family,
  covariate_layers, fe_equation, centre_scale = T, time_var = "year",
  time_scale = c(2000, 2005, 2010, 2015), weight_col = NULL,
  num_fold_cols = 1, num_folds = 1, cores = 1, sge_parameters = NULL)

Arguments

...	Initialized models. If blank, default versions of earth and gam are created.
inlist	logical. Are the models being passed through ... already in a list format?
data	data table. Dataset to be machine learned.
indicator	character vector. Name of the indicator (and by extension) the column name of the dependant variable
indicator_family	character vector. Designates the statistical family that should be modeled. Usually 'binomial' or 'gaussian'
covariate_layers	list of raster like objects. A named list of raster like objects of covariates
fe_equation	character vector of an equation. The equation specifying the fixed effects portion of the model. It should match with the names of covariate_layers.
centre_scale	logical. Determines whether the covariate values are centered/normalized before being returned. Binary variables are ignored.
time_var	character vector. Name of the column denoting the time (e.g. period or year) of a given data point
time_scale	numeric vector. List of years or times that the time var correlates to.
weight_col	character vector. Denotes the column (if applicable) in the dataset that specifies the data weights
num_fold_cols	numeric or character. Number of columns/interations for crossfold validation. if a character string, assume it refers to columns already existing in data. They will be renamed to sfold_#
num_folds	numeric. The number of folds the data is split on.
cores	numeric. The number of cores available for parallel computation
sge_parameters	object returned from init_sge. Provides sge parameters to govern submodel computation. If NULL, mclapply is used to run submodels instead

Value

Stacker governor object

dahcase/mbgstacking documentation built on May 20, 2019, 4:08 p.m.