Skip to contents

Gibbs refresh of the latent normal matrix Z for the censored-binary (cbin) family, where each actor emits a capped set of binary nominations. Within every sending row the update honours three order constraints: a nominated tie has a positive latent value, every nominated tie dominates every non-nominated tie, and a non-nominated tie in a row that has not spent its full nomination budget (odobs < odmax) is held below zero. Reciprocal cells are coupled through the dyadic correlation rho; the self-loop diagonal and any missing cells are redrawn from their unconstrained conditionals.

Usage

rZ_cbin_fc(Z, EZ, rho, Y, odmax, odobs)

Arguments

Z

current latent matrix (square).

EZ

conditional mean of Z.

rho

dyadic (within-dyad) correlation.

Y

observed nomination matrix, entries in {0, 1} with NA for unobserved cells.

odmax

nomination cap, a scalar or length-nrow(Z) vector.

odobs

observed out-degree per actor.

Value

the updated latent matrix, same shape and dimnames as Z.

Author

lame authors