analyseMCMC_stats.f90 File Reference

(AnalyseMCMC version 0.7.1, revision 448, hash 2f54722 (2013-11-19))

Compute statistics for analyseMCMC. More...

Functions/Subroutines
subroutine	statistics (exitcode)
	Compute basic statistics for analyseMCMC. More...
subroutine	save_stats (exitcode)
	Determine the number of points to use for the computation of R-hat Since we need a fixed number of points for all chains, take nn = the minimum of (the length of each chain after the burn-in) and use the last nn data points of each chain. More...
subroutine	save_bayes (exitcode)
	Save Bayes-factor statistics to file. More...
subroutine	save_cbc_wiki_data (ic)
	Save data in CBC wiki format. More...
subroutine	compute_mixing (mynn, print_data)
	Compute mixing (R-hat) for multiple chains. Use unwrapped data. More...
subroutine	compute_autocorrelations ()
	Determine the number of points to use for the computation of R-hat Since we need a fixed number of points for all chains, take nn = the minimum of (the length of each chain after the burn-in) and use the last nn data points of each chain. More...

Detailed Description

Compute statistics for analyseMCMC.

Function/Subroutine Documentation

subroutine statistics

(

integer, intent(out)

exitcode

)

Compute basic statistics for analyseMCMC.

Return values

exitcode

Exit status code

• Other results are returned using a number of modules

References compute_autocorrelations(), compute_mixing(), revper(), save_cbc_wiki_data(), and set_derivedparameternames().

Referenced by analysemcmc().

subroutine save_stats

(

integer, intent(out)

exitcode

)

Determine the number of points to use for the computation of R-hat Since we need a fixed number of points for all chains, take nn = the minimum of (the length of each chain after the burn-in) and use the last nn data points of each chain.

Save statistics to file

Return values

exitcode

Exit status code

Referenced by analysemcmc().

subroutine save_bayes

(

integer, intent(out)

exitcode

)

Save Bayes-factor statistics to file.

Return values

exitcode

Exit status code

Referenced by analysemcmc().

subroutine save_cbc_wiki_data

(

integer, intent(in)

ic

)

Save data in CBC wiki format.

Parameters

ic	Chain number

References rev2pi(), and rrevpi().

Referenced by statistics().

subroutine compute_mixing	(	integer, intent(in)	mynn,
		logical, intent(in)	print_data
	)

Compute mixing (R-hat) for multiple chains. Use unwrapped data.

Parameters

mynn	Number of iterations to compute Rhat for. Use nn=0 for full chains. The last nn data points are used.
print_data	Print results or not.

Note

This works only for fixed chain length, so take the min N

See Also

Brooks & Gelman, Journal of Computational and Graphical Statistics, Vol.7, Nr.4, p.434-455, 1998:

• "General Methods for Monitoring Convergence of Iterative Simulations"
• http://www.jstor.org/pss/1390675 (for purchase)
• http://www.stat.columbia.edu/~gelman/research/published/brooksgelman.pdf (Author's website)
• See Eq.1.1, where: B/n := meanVar and W := chVar

Referenced by statistics().

subroutine compute_autocorrelations

(

)

Determine the number of points to use for the computation of R-hat Since we need a fixed number of points for all chains, take nn = the minimum of (the length of each chain after the burn-in) and use the last nn data points of each chain.

Compute autocorrelations

• use the original nChains0 chains and allDat()
• results are printed to stdout if prAcorr>0

Referenced by statistics().