moving parentheses in network.extract.with.lasttoggle

[sgoodreau]

in the lines extracting vectors of tails and heads.
The issue was identified by Lin Zhu, and explained in an email exchange.
I have dived into the code to understand the logic. I agree that her edits are logically correct. I have tested with and without the edits and confirmed that the edits work.

Code below to demonstrate:

# load packages, all LATEST versions as of 2025.10.07-------

library(ergm)                # 4.9.0
library(tergm)               # 4.2.2
library(network)             # 1.19.0
library(networkDynamic)      # 0.11.5
library(sna)                 # 2.8
library(tsna)                # 0.3.6
library(statnet.common)      # 4.12.0

# clear env
rm(list=ls())
set.seed(1)

# create networkDynamic object
nw <- network(100, directed = FALSE)
nwd <- as.networkDynamic(nw) 
nwd <- activate.vertices(nwd, onset=0, terminus = Inf)
nwd <- activate.edges(nwd, onset = 0, terminus = Inf)
set.network.attribute(nwd,'vertex.pid','vertex.names')

# fit model
myergm <- ergm(nwd~edges + degree(1), target.stats = c(150, 30)) 

# adjust formation coef
myergm_adj <- myergm
myergm_adj$coefficients[1]  <- coef(myergm)[1] - 3.555348
  
# fit model without deleting node
at <- 2
my_sim_1 <- simulate(nwd,
                seed = 12345,
                formation = ~edges + degree(1),
                dissolution = ~offset(edges),
                coef.form = coef(myergm_adj),
                coef.diss = 3.555348,
                time.start = at,
                time.slices = 10,
                time.offset = 0,
                control = control.simulate.network(MCMC.burnin.min = 1e5))

# delete one vertex

nwd_del <- nwd
nwd_del <- delete.vertices(nwd_del, 1)

# fit model on network with one vertex deleted
at <- 2
my_sim_del <- simulate(nwd_del,
                     seed = 12345,
                     formation = ~edges + degree(1),
                     dissolution = ~offset(edges),
                     coef.form = coef(myergm_adj),
                     coef.diss = 3.555348,
                     time.start = at,
                     time.slices = 10,
                     time.offset = 0,
                     control = control.simulate.network(MCMC.burnin.min = 1e5))

# without fix, returns error: Error in lttails[w2] <- ltheads[w2] : 
#      NAs are not allowed in subscripted assignments

# traceback:
# 11: network.extract.with.lasttoggle(object, start)
# etc.

# Inside network.extract.with.lasttoggle:
# line 4: lts <- unlist(lapply(lapply(lapply(nwd$mel, "[[", "atl"), 
#                  "[[", "active"), function(x) suppressWarnings(max(x[x <= 
#                  at]))))[valid_eids]
# yields:
# [1] 80 25 16 71 27  2 57 43 57  6 14  3 20 53 19  7  9  4 56 34 35 27 74 15 24 45 22 30
# [29] 16 67 59 74 64 13 28 23 29 34 18  5 11 12 20 53 18  4 12 52 12 21 38 55 91 14  5  2
# [57] 40 49 16  9 75 10 39  3  7 27 12 45 55 18  5 51  5 41 50 48 21 20 14 75 22  3 29 60
# [85]  3 11  3 14 13 20 34 64  7  2 NA NA NA NA

# the NAs are because the code in network.extract.with.lasttoggle says:
#   tails <- unlist(lapply(nwd$mel, "[[", "outl"))[valid_eids]
#   heads <- unlist(lapply(nwd$mel, "[[", "inl"))[valid_eids]
# and this code first pulls out the entries, then checks for valid_eids, then unlists.  
# But it should check for valid_ids, then pull out the entries, then unlist.  I.e.:
#   tails <- unlist(lapply(nwd$mel, "[[", "outl")[valid_eids])
#   heads <- unlist(lapply(nwd$mel, "[[", "inl")[valid_eids])
# It appears that the code might actually also work as:
#   tails <- unlist(lapply(nwd$mel, "[[", "outl"))
#   heads <- unlist(lapply(nwd$mel, "[[", "inl"))
# because during the lapply any with NULL values get dropped anyway, so there's 
#   no need to re-drop them by using [valid_eids]. But the first version seems safer.

[krivit]

Thanks!