Internal class for a binary SBM fit under MAR sampling designs with covariates.

Super classes

sbm::SBM -> sbm::SimpleSBM -> SimpleSBM_fit -> SimpleSBM_fit_withCov

Active bindings

imputation: the matrix of imputed values
vExpec: double: variational approximation of the expectation complete log-likelihood
vExpec_corrected: double: variational approximation of the expectation complete log-likelihood with correction to be comparable with MNAR criteria

Methods

Public methods

SimpleSBM_fit_withCov$update_parameters()
SimpleSBM_fit_withCov$update_blocks()
SimpleSBM_fit_withCov$polish_log_tau()
SimpleSBM_fit_withCov$clone()

Inherited methods

`SimpleSBM_fit_withCov$update_parameters()`

update parameters estimation (M-step) via Newton-Raphson: the M-step objective is a weighted logistic regression (concave), so Newton converges in a handful of iterations – no external optimizer is required.

Usage

SimpleSBM_fit_withCov$update_parameters(...)

Arguments

...: use for compatibility

`SimpleSBM_fit_withCov$update_blocks()`

update variational estimation of blocks (VE-step)

Usage

SimpleSBM_fit_withCov$update_blocks(...)

Arguments

...: use for compatibility

`SimpleSBM_fit_withCov$polish_log_tau()`

for each node, the complete-data log-likelihood it would contribute to each class if hard-assigned there (theta/beta/pi held fixed), used to decide node-swap moves in missSBM_fit$polish().

Usage

SimpleSBM_fit_withCov$polish_log_tau(log_lambda = 0)

Arguments

log_lambda: additional sampling-design-dependent term, added as-is

Returns

an N x Q matrix

`SimpleSBM_fit_withCov$clone()`

The objects of this class are cloneable with this method.

Usage

SimpleSBM_fit_withCov$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.