Skip to content

Latest commit

 

History

History
41 lines (27 loc) · 2.18 KB

AMCMC_appendix.md

File metadata and controls

41 lines (27 loc) · 2.18 KB
Raw

Description

This is an appendix to AMCMCinfo.txt. Here we describe additional functions that are provided with Adaptive_Gibbs.cpp:

set_working_directory(String directory) 

String get_working_directory()

print_working_directory()

NumericVector trace_coord(int coord)

NumericMatrix trace_weights()

NumericMatrix trace_proposals()

NumericVector  trace_inv_sp_gap()

NumericMatrix estimated_covariance()

env Rpointer(input)

Arguments

  • directory - string, specifying a path where the output of the chain should be stored. It is recommended to specify the directory, e.g., directory = "~/AdaptiveGibbs/simulation_results". By default, directory = "", meaning that all the output is written to a root folder;
  • coord - integer specifying a coordinate for which its' trace should be extracted;
  • input - virtually any R object;

Details:

  • set_working_directory(directory) - assigns directory to the paths where all the output is stored, which we call working_directory;
  • get_working_directory() - returns a string of the working_directory;
  • print_working_directory() - prints the working_directory;
  • trace_coord(coord) - returns a vector of a trace of a coordinate coord;
  • trace_weights() - returns a matrix each row of which is a trace of adapted probability weights as in the Adaptive Random Scan Gibbs Sampler from the ARSG paper;
  • trace_proposals() - returns a matrix each row of which is a trace of adapted proposal coefficients. Here the coefficients are tuned in a way that retains the average acceptance ratio around 0.44 for one-dimensional updates, and around 0.234 for multi-dimensional updates;
  • trace_inv_sp_gap() - returns a vector of a trace of the estimated value for 1/pseudo-spectral gap discussed in the ARSG paper;
  • estimated_covariance() - returns estimated covariance matrix;
  • Rpointer(input) - creates an analogue of C-pointer to input. The returned object has $value attribute which returns the value of Rpointer. This function should be used if R-defined density has external parameters that need to be changed during the algorithm run. Refer to the tutorial_examples.R for an example of usage.