Learn R Programming

lame (version 1.3.4)

rZ_frn_bip_fc: Sample Z under bipartite fixed-rank nominations (rectangular)

Description

Rectangular analogue of rZ_frn_fc. Row-wise rank constraints: ranked nominees are positive and ordered by rank; non-nominees fall below all ranked cells in the row (and below 0 when the row is unsaturated). With rho = 0 the unipartite triangle coupling drops out.

Usage

rZ_frn_bip_fc(Z, EZ, Y, YL, odmax, odobs)

Value

updated nA x nB latent matrix.

Arguments

Z

current rectangular latent matrix (nA x nB).

EZ

expected value (nA x nB).

Y

observed rank matrix (nA x nB); 0 = unranked, 1..ncol(YL) = rank order.

YL

per-row ranked-receiver list (nA x max_rank).

odmax

row-wise max nominations.

odobs

row-wise observed outdegree.

Author

Cassy Dorff, Shahryar Minhas, Tosin Salau

Details

Monotonicity convention (important for simulation). The sampler enforces higher Y -> higher Z: when Y[i, j] > Y[i, k], the corresponding latent values satisfy Z[i, j] > Z[i, k]. When simulating data for recovery tests, build Y via rank(Z[i, ]) (low rank = small Y), not rank(-Z[i, ]). the inverted convention identifies the negative of the true beta.