BayesFactorSC: Bayes Factor Approximation using Schwarz Criterion

In audreyqyfu/dagr: Generate and analyze data for direced acyclic graphs

Description

Approximates Bayes Factor using Schwarz Criterion. The function uses BIC to calculate Schwarz Criterion.

Usage

BayesFactorSC(firstModel = 'model0',
              secondModel = 'model1',
              firstMHOutput,
              secondMHOutput,
              data,
              log = FALSE)

Arguments

firstModel	String indicating the model that will be in the numerator of Bayes Factor.
secondModel	String indicating the model that will be in the denominator of Bayes Factor.
firstMHOutput	Markov chain matrix of the parameters from the model specified in the firstModel argument. This is the first element in the list of the output from the MHSampler function.
secondMHOutput	Markov chain matrix of the parameters from the model specified in the secondModel argument. This is the first element in the list of the output from the MHSampler function.
data	Data frame. First column must be the gene expression data for gene one, the second column must be the gene expression data for gene two, and the third column must be the genotype data.
log	Logical argument. If TRUE the Bayes Factor approximation will be returned on the log scale.

Value

A real number. The scale depends on the input of the log argument.

References

Robert E. Kass and Adrian E. Raftery. Bayes Factors. Journal of the American Statistical Association, Vol.90, 773-795, 1995.

Examples

# Values used to simulate the data for model 3.
actualValues <- list(b0.1 = 3.2, b1.1 = 1.7, b0.2 = 4.6, b1.2 = 3.1, sd.1 = 2.4, sd.2 = 0.9, rho = 0.72)

# Number of iterations to run the MHSampler function.
m <- 5000

# Simulate data under model 3
Model3 <- simulateData(N = 100,
                       p = 0.43,
                       seed = 338,
                       model = 'model3',
                       parameters = actualValues)

# Run the MHSampler function for models 3 and 4 and calculate Bayes Factor for the two models.
Output3 <- MHSampler(nIterations = m,
                     burnIn = 0.2,
                     thinned = 200,
                     model = 'model3',
                     data = Model3,
                     parameters = list(b0.1 = 0, b1.1 = 1, b0.2 = 0, b1.2 = 1, sd.1 = 1, sd.2 = 1, rho = 0.72),
                     priorParameters = list(b0.1 = c(0, 1), b1.1 = c(0, 1), b0.2 = c(0, 1), b1.2 = c(0, 1), sd.1 = c(1, 1), sd.2 = c(1, 1), rho = c(0, 0.5)),
                     proposalSD = list(b0.1 = 1, b1.1 = 1, b0.2 = 1, b1.2 = 1, sd.1 = 1, sd.2 = 1, rho = 0.5))

Output4 <- MHSampler(nIterations = m,
                     burnIn = 0.2,
                     thinned = 200,
                     model = 'model4',
                     data = Model3,
                     parameters = list(b0.1 = 0, b1.1 = 1, b0.2 = 0, b1.2 = 1, sd.1 = 1, sd.2 = 1, rho = 0.72),
                     priorParameters = list(b0.1 = c(0, 1), b1.1 = c(0, 1), b0.2 = c(0, 1), b1.2 = c(0, 1), sd.1 = c(1, 1), sd.2 = c(1, 1), rho = c(0, 0.5)),
                     proposalSD = list(b0.1 = 1, b1.1 = 1, b0.2 = 1, b1.2 = 1, sd.1 = 1, sd.2 = 1, rho = 0.5))

BFA <- BayesFactorSC(firstModel = 'model3',
                     secondModel = 'model4',
                     firstMHOutput = Output3[[1]],
                     secondMHOutput = Output4[[1]],
                     data = Model3,
                     log = FALSE)

audreyqyfu/dagr documentation built on May 20, 2019, 5:05 p.m.