build_cindep_model: Build a multivariate, conditionally independent model using...

In MichaelHoltonPrice/yada: Yet Another Demographic Analysis package

Description

Build a multivariate, conditionally independent model using the univariate cross-validation results. Optionally, the standard error of the unconstrained parameter vector (th_y_bar) is calculated to support the re-scaling done in fit_multivariate.

Usage

build_cindep_model(
  data_dir,
  analysis_name,
  fold = NA,
  calc_se = FALSE,
  save_file = F,
  allow_corner = FALSE
)

Arguments

data_dir	Directory to save univariate model parameters
analysis_name	analysis_name for file-naming
fold	Fold number. If NA, build model from main solution.
calc_se	Whether to calculate the standard error for th_y_bar
save_file	Logical whether to save cindep model as an .rds file

Value

A list with the parameter vector (th_y), model specification (mod_spec), and transformed parameter vector (th_y_bar). If calc_se is TRUE, the standard error for th_y_bar is added (th_y_bar_se).

(Default: FALSE) If FALSE, an error is thrown if any standard errors have negative values, which usually happens because an optimum is a corner solution. If TRUE, an attempt is made to use pertinent median values from other variables to set the scale.

MichaelHoltonPrice/yada documentation built on Sept. 19, 2021, 11:27 p.m.