gforce.glatent_confints: Confidence Intervals for Estimation in G-Latent Models.

Description

Estimate the precision matrix and construct confidence intervals for the latent and group averages graphs. The user can either provide an estimate of the relevant covariance matrix or the data and cluster assignments. If cross validation is selected, the data and clusters must be provided.

Usage

gforce.glatent_confints(C_hat = NULL, X_vals = NULL, clusters = NULL,
  alpha = 0.05, graph = "latent", variance_estimator = "simple",
  use_cv = FALSE, cv_opts = NULL, lambda1 = NULL, lambda2 = NULL)

Arguments

C_hat

a \(d x d\) matrix. This is the estimated latent or group averages covariance matrix.

X_vals

a \(n x d\) matrix. This is a matrix where each row is a sample from the model.

clusters

a \(d\) dimensional integer vector. Contains the assignment of variables to groups.

alpha

a numeric. alpha is the confidence level.

graph

a string. It can either have value 'latent' or 'averages'.

variance_estimator

a string. It can either have value 'simple' or 'exact'.

use_cv

logical expression. Indicates whether or not to use cross validation.

cv_opts

an object. Contains options for cross validation procedure.

lambda1

a numeric. Parameter for the first optimization problem.

lambda2

a numeric. Parameter for the second optimization problem.

Value

a \(d x d x 3\) array. The first \(d x d\) slice is the lower confidence bound, the second the point estimate, and the third the upper confidence bound.

References

C. Eisenach, F. Bunea, Y. Ning, and C. Dinicu. Efficient, High-Dimensional Inference for Cluster-Based Graphical Models. Manuscript submitted for publication, 2018.

Examples

Run this code

# NOT RUN {
K <- 5
n <- 50
d <- 50
dat <- gforce.generator(K,d,n,3,graph='scalefree')
th_tilde <- gforce.glatent_confints(X_vals = dat$X,clusters = dat$group_assignments,
                                    use_cv = TRUE,graph='latent')

# }