BGGM (version 1.0.0)

select.explore: Select Graphical Structure with the Bayes Factor

Description

This allows for not only estimating the conditional dependence structure, that is non-zero edges, but also the conditional independence structure (evidence for no relation).

Usage

# S3 method for explore
select(object, BF_cut = 3, alternative = "two.sided", ...)

Arguments

object

object of class explore.default

BF_cut

evidentiary threshold

alternative

type of hypothesis (see notes)

...

currently not used

Value

list of class select.explore:

alternative = "two.sided":

• partials_non_zero selected partial correlation matrix

• pcor_mat partial correlation matrix (non set to zero)

• pcor_sd partial correlation standard deviations

• Adj_10 adjacency matrix for the selected edges (in favor of the alternative)

• Adj_01 adjacency matrix for the null hypothesis (conditional independence)

• BF_10 Bayes factors for alternative

• BF_01 Bayes factors for the null hypothesis

• BF_cut evidentiary threshold

• alternative "two.sided"

• code match.call()

alternative = "greater":

• partials_positive selected partial correlation matrix

• pcor_mat partial correlation matrix (none set to zero)

• pcor_sd partial correlation standard deviations

• Adj_20 adjacency matrix for the selected edges (in favor of the alternative)

• Adj_01 adjacency matrix for the null hypothesis (conditional independence)

• BF_20 Bayes factors for alternative

• BF_01 Bayes factors for the null hypothesis

• BF_cut evidentiary threshold

• alternative "greater"

• code match.call()

alternative = "less":

• partials_negative selected partial correlation matrix

• pcor_mat partial correlation matrix (none set to zero)

• pcor_sd partial correlation standard deviations

• Adj_20 adjacency matrix for the selected edges (in favor of the alternative)

• Adj_01 adjacency matrix for the null hypothesis (conditional independence)

• BF_20 Bayes factors for alternative

• BF_01 Bayes factors for the null hypothesis

• BF_cut evidentiary threshold

• alternative "less"

• code match.call()

alternative = "exhaustive"

• post_prob data.frame with posterior probabilities for each edge

• neg_mat adjacency matrix for negative edges

• post_mat adjacency matrix for positive edges

• null_mat adjacency matrix for zero (conditional independence)

• "alternative" "exhaustive"

• pcor_mat partial correlation matrix (non set to zero)

• pcor_sd partial correlation standard deviations

• code match.call()

• prob hyp_prob

Examples

# NOT RUN {
# p = 10
Y <- BGGM::bfi[,1:10]

# sample posterior
fit <- explore(Y, iter = 5000)

# select E
E <- select(fit, BF_cut = 3)

# summarize
summary(E)

# non-zero edges
E$partials_non_zero

# adjacency matrix
E$Adj_10

# null adjacency matrix
E$Adj_01
# }