fitBabyMonitor: Fit Baby-MONITOR for CPQCC/VON

In dhelkey/dgrank: Construct Draper-Gittoes Rankings of Institutions

Description

fitBabyMonitor comprehensively applies the Baby-MONITOR score Returns a large list. Use inst_mat for institution rankings, full_subset_mat_baseline and full_subset_mat_nobaseline for subset rankings, and subset_baseline_mat and subset_nobaseline_mat for subset rankings within institution.

Usage

fitBabyMonitor(minimal_data, num_cat, num_cont, var_intercept, var_inst,
  var_cat, var_cat_2way, var_cont, var_subset = 1, var_inst_subset_2way = 1,
  subset = FALSE, fit_method = "probit", sparse = TRUE, burn_in = 100,
  iters = 1000, alpha = 0.05, verbose = TRUE)

Arguments

minimal_data	Data_frame with a particular format: 1st column: Outcome vector (0-1 encoding) 2nd column: Institution ID 3rd column (if subset == TRUE): Next: num_cat columns of categorical variables (num_cat can equal 0) Next: num_cont columns of continuous variables (num_cont can equal 0)
num_cat	Scalar number of categorical variables.
num_cont	Number of continuous variables.
var_intercept	Prior variance for intercept parameter.
var_inst	Prior variance for institution parameters.
var_cat	Prior variance for categorical parameters.
var_cat_2way	Prior variance for categorical interaction parameters.
var_cont	Prior variance for continuous parameters.
var_subset	Prior variance for subset parameters.
var_inst_subset_2way	Prior variance for subset interaction parameters.
subset	If TRUE, perform subset fitting tasks
fit_method	Analysis method. Options: 'probit' for Probit Regression, 'bayesianregression' for Bayesian linear regression, and 'probitk' for a Probit Regression implementation written by Daniel Kirsner.
sparse	Should design_matrix be stored as Sparse matrix? Requires Matrix package.
burn_in	Number of initial iterations to discard for burn in.
iters	Number of MCMC iterations to use.
alpha	We look at posterior (1-alpha)% posterior intervals.
verbose	If TRUE, display status messages while fitting

Value

Returns a large list with the following components:

inst_mat: Matrix (one row per institution) computing summary statistics, D-G institution ranking, and intervals

full_subset_mat_baseline, full_subset_mat_nobaseline: Matrix with rankings and intervals for the various subset categories.

subset_baseline_mat, subset_nobaseline_mat: Matrix with rankings and intervals for subset categories within institution.

group_labels: A vector of the names of each institution

mcmc_fit: Matrix of MCMC iterations for each coefficient dg_z: Matrix of computed z score for each institution at each MCMC iterations.

coefs: Names of each coefficient (1st is intercept, then institution, then everything else)

prior_var_vec: Vector of prior variances for each coefficient

model_matrix: Design matrix

dhelkey/dgrank documentation built on May 12, 2019, 5:24 p.m.