Skip to contents

Convert a netify object to the format expected by lame::ame()

Source: R/to_lame.R
netify_to_lame.Rd

netify_to_lame() (also available as to_lame()) is a thin specialization of \link{netify_to_amen} for the optional lame workflow. lame::ame() accepts the same y / xdyad / xrow / xcol skeleton as amen::ame() but adds two things netify users care about:

Usage

netify_to_lame(
  netlet,
  lame = FALSE,
  family = NULL,
  pad = TRUE,
  fit_method = c("gibbs", "als"),
  bootstrap = 0L
)

to_lame(
  netlet,
  lame = FALSE,
  family = NULL,
  pad = TRUE,
  fit_method = c("gibbs", "als"),
  bootstrap = 0L
)

Arguments

netlet

a netify object.

lame

logical. as in netify_to_amen: pass TRUE when actor composition varies over time. default FALSE.

family

optional character. ame family to use. if NULL (default), inferred from the netify: "binary" for binary networks, "normal" for weighted, with a once-per-session info message naming the choice.

pad

logical. when lame = TRUE and the per-period list is returned (the function never actually pads in place; it always returns a list), controls whether the once-per-session info message points users at the lame::lame() call baked into nl$ame_call. default TRUE (emit the message); set FALSE to silence it.

fit_method

one of "gibbs" (default – mcmc posterior via lame::ame() / lame::lame()) or "als" (fast alternating- least-squares point estimate via lame::ame_als()). for bipartite + binary networks where mcmc is slow, als with bootstrap > 0 gives a fast point estimate plus parametric/block bootstrap uncertainty intervals in a single call. the choice only affects the generated ame_call snippet – the y/xdyad/ xrow/xcol payload is identical.

bootstrap

integer. number of bootstrap replicates for als uncertainty intervals. ignored when fit_method = "gibbs" (mcmc draws are the uncertainty representation there). default 0 (no bootstrap). pass 200 for a reasonable interval estimate.

Value

a list with the same shape as netify_to_amen output, plus two helpful extras:

mode

character: "unipartite" or "bipartite". pass this directly to lame::ame(mode = .).

family

character: the suggested family ("binary" or "normal"). pass to lame::ame(family = .).

ame_call

character: a literal lame::ame() call (cross-sectional) or lame::lame() snippet (longitudinal) the user can copy-paste.

Details

  • rectangular y for bipartite networks via mode = "bipartite" – the standard amen::ame() rejects this.

  • ragged longitudinal panels via lame::lame(), which aligns per-period actor sets by name – amen requires constant actor composition.

this function wraps \link{netify_to_amen} and:

  1. picks a family default appropriate to the netify ("binary" for binary nets, "normal" for weighted),

  2. suggests the correct mode argument for lame::ame() ("unipartite" / "bipartite"),

  3. for longitudinal output, emits a copy-paste lame::lame() call in the returned ame_call slot. with lame = TRUE the per-period lists are passed directly; with lame = FALSE the snippet first splits the [n, n, t] arrays into the per-period lists that lame::lame() expects.

bipartite + binary: the case to_amen() cannot fit. for a bipartite weighted-binary netify (e.g., person x event attendance, 0/1), this is the exact pipeline:


bp <- netify(df, actor1 = "person", actor2 = "event",
mode = "bipartite", weight = "attended")
nl <- to_lame(bp) # auto-detects binary
fit <- lame::ame(
Y = nl$Y, Xrow = nl$Xrow, Xcol = nl$Xcol,
mode = nl$mode, family = nl$family,
nscan = 1000, burn = 500
)
# uncertainty: posterior intervals are in fit$BETA / fit$VC

ragged longitudinal panels. when lame = TRUE and actors enter / exit the network, this function:

  1. builds the per-period list via netify_to_amen(lame=TRUE),

  2. bakes a lame::lame() call into nl$ame_call that passes the per-period lists directly; lame::lame() aligns actors across periods by name, for unipartite and bipartite networks.

the returned Y / Xdyad / Xrow / Xcol are always the per-period list – pad only controls whether the once-per- session info message reminding the user of the snippet fires.

See also

netify_to_amen (the underlying converter), netify_to_statnet, netify_to_dbn, netify_to_igraph.

Author

shahryar minhas

cassy dorff, shahryar minhas