MRH (version 2.2)

DIC: DIC CALCULATES INFORMATION CRITERION GIVEN THE MCMC CHAINS FROM THE estimateMRH routine.

Description

The DIC function calculates the Deviance Information Criterion given the MCMC chains from an estimateMRH routine, using the formula: DIC = .5*var(D)+mean(D), where D is the chain of -2*log(L), calculated at each retained iteration of the MCMC routine. It also provides the Aikaike Information Criterion (AIC) = 2*p + D and the Bayesian Information Criterion (BIC) = p*ln(n) + D, where `p' is the number of parameters in the model. Both AIC and BIC report the maximum (i.e. "worst") values in the chain.

Usage

DIC(mrhobject, n)

Arguments

mrhobject

The chains found in the MCMCchains.txt file, created using the estimateMRH routine, or the MRH results object.

n

The sample size of the original dataset. If n is not entered, the BIC calculation will not be returned to the user.

Value

loglik.summ

The summary of the chain of -2*loglike values.

ICtable

Table containing the DIC, AIC and BIC values.

Details

The number of parameters `p' is calculated as 2^M (one for each split parameter Rmp, and one for the cumulative hazard at H), plus 2 for a and lambda (parameters in the Gamma prior for H), and one for each covariate included under the proportional hazards assumption. If k and/or gamma are sampled (parameters in the prior for Rmp), the number of estimated parameters is increased by 1 for k and 2^M-1 for gamma. If a covariate is included under the non-proportional hazards assumption, the number of estimated parameters (excluding any covariates included under the proportional hazards assumption) is multiplied by the number of strata in the non-proportional covariate.

References

Gelman, A., Carlin, J., Stern, H., Dunson, D., Vehtari, A., Rubin, D. (2004) Bayesian Data Analysis.

Spiegelhalter, D.J., Best, N.G., Carlin, B.P., van der Linde, A. (2002), Bayesian measures of model complexity and fit (with discussion). Journal of the Royal Statistical Society, Series B. 64, 583--639.

Examples

# These MRH fit portion of the examples 
# are from the estimateMRH() help page.  
# They do not need to be re-run if the 
# objects are already in the active workspace.
data(cancer)
cancer$censorvar = cancer$status - 1
## Not run: 
# fit.lung = estimateMRH(formula = Surv(time, censorvar) ~ 
# 	age + as.factor(sex) + ph.karno, data = cancer,
# 	M = 3, maxStudyTime = 960, burnIn = 200, maxIter = 1000, 
# 	thin = 1, outfolder = 'MRH_lung')
# 
# fit.lung.prune = estimateMRH(formula = Surv(time, censorvar) ~ 
# 	age + as.factor(sex) + ph.karno, data = cancer,
# 	M = 3, maxStudyTime = 960, burnIn = 200, maxIter = 1000, 
# 	thin = 1, prune = TRUE, outfolder = 'MRH_lung_prune')## End(Not run)
	
# Compare the DIC of the pruned and unpruned models.
# The sample size must be entered for calculation of BIC.
# This number can be found in the ``MCMCInfo.txt" file
# in the output folder.
## Not run: 
# DIC(fit.lung, n = 227)
# DIC(fit.lung.prune, n = 227)## End(Not run)