fit_multivariate: Fit a multivariate mixed cumulative probit model.

In MichaelHoltonPrice/yada: Yet Another Demographic Analysis package

Description

Do a maximum likelihood fit of multivariate mixed cumulative probit model using the Hooke-Jeeves (hjk) algorithm. The parameter vector used in the optimization has, roughly, a mean of zero and an expectation that the best actual value is within about +/-1 of the starting value. This is accomplished by subtracting the mean of the conditionally dependent fit and dividing by the standard error of the conditionally dependent fit.

Usage

fit_multivariate(
  x,
  Y,
  mod_spec,
  cindep_model,
  save_file = NA,
  hjk_control = list()
)

Arguments

x	The vector of independent variables
Y	The matrix of responses
mod_spec	The model specification
cindep_model	A list specifying the conditionally independent model (likely the output of build_cindep_model). cindep_model must contain cindep_model$th_y_bar and cindep_model$th_y_bar_scale.

Value

A list object containing the best-fit parameter vector (th_y), and the unconstrained best-fit parameter (th_y_bar), and the return object from the optimization (hjk_output).

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