Error with BYM2 in inlabru marked LGCP
Morgane Declerck
Dear INLA Discussion Group,
I am fitting a marked LGCP using inlabru + INLA (Cox process + binomial mark model).
The model runs successfully when using a cyclic generic0 effect for tide-varying slopes (based on a custom precision matrix). However, when I replace this with a BYM2 model, the fit fails with the following error:
*** Please (re-)consider your model, priors, confounding, etc.
***[0] thread_num[0] warning *** iterative process seems to diverge, 'vb.correction' is aborted
(repeated several times)
Error in `data.frame()`:
! arguments imply differing number of rows: 1, 0
Backtrace:
base::system.time(...)
inlabru::bru(...)
inlabru::iinla(...)
base::data.frame(...)
Model setup: BYM2 version that fails
# ============================================================
# 6) BYM2 CYCLIC TIDE GRAPH
# ============================================================
# Tide groups are treated as cyclic neighbours:
# 1 -- 2 -- 3 -- 4 -- 1
make_tide_graph <- function() {
graph_file <- tempfile(fileext = ".adj")
cat(
"4\n",
"1 2 2 4\n",
"2 2 1 3\n",
"3 2 2 4\n",
"4 2 3 1\n",
file = graph_file,
sep = ""
)
INLA::inla.read.graph(graph_file)
}
tide_graph <- make_tide_graph()
# ------------------------------------------------------------
# Outer boundary (km units)
# ------------------------------------------------------------
outer_boundary <- fm_nonconvex_hull(domain_km2, 40)
# ------------------------------------------------------------
# HEXAGONAL LATTICE MESH
# ------------------------------------------------------------
mesh <- fm_mesh_2d_inla(
loc = fm_hexagon_lattice(domain_km2, edge_len = 80),
boundary = list(domain_km2, outer_boundary),
max.edge = c(90, 150),
cutoff = 70,
crs = crs_km
)
spde <- inla.spde2.pcmatern(mesh = mesh,
prior.range = c(150, 0.5),
prior.sigma = c(1, 0.05)
)
cmp <- ~ -1 +
point_field(geometry, model = spde) +
log_PEA(
tide_group_id,log_PEA_z,model = "bym2",
graph = tide_graph,
scale.model = TRUE,
constr = TRUE
) +
log_surface_chla(tide_group_id, log_surface_chla_z,
model = "bym2",
graph = tide_graph,
scale.model = TRUE,
constr = TRUE
) +
sst_effect(tide_group_id,sst_z,
model = "bym2",
graph = tide_graph,
scale.model = TRUE,
constr = TRUE
) +
sal_effect(tide_group_id,sal_depth_int_z,
model = "bym2",
graph = tide_graph,
scale.model = TRUE,
constr = TRUE
) +
Inter_point(1) +
Inter_mark(1) +
beta_link(
main = 1,
model = "linear",
mean.linear = 0,
prec.linear = 1,
n = 1,
values = 1
) +
log_dist_col_m_z + indiv(UniqueID, model = "iid") +
mark_field_tide(geometry, model = spde, replicate = tide_group_id)
obs_points <- bru_obs(geometry ~ Inter_point + point_field,
family = "cp",
data = gan_km,
domain = list(geometry = mesh),
samplers = list(geometry = domain_km2),
tag = "points"
)
mark_formula <- mark ~
Inter_mark + log_PEA +log_surface_chla +sst_effect +sal_effect +log_dist_col_m_z + indiv +
point_field * beta_link +
mark_field_tide
obs_marks <- bru_obs(mark_formula,family = "binomial",data = gan_km,tag = "marks")
fit <- bru(
components = cmp,obs_points, obs_marks,
options = list(
control.compute = list(
dic = TRUE,
waic = TRUE,
cpo = TRUE,
config = FALSE
),
control.predictor = list(compute = TRUE),
control.inla = list(
strategy = "simplified.laplace",
int.strategy = "auto"
),
bru_verbose = TRUE
)
)
Any guidance on how to resolve this error would be greatly appreciated.
The variable tide_group_id has 4 levels with cyclic adjacency (1–2–3–4–1).
Is BYM2 appropriate for such a small graph (4 nodes)?
Would generic0 be more suitable for cyclic effects with few levels?
Thank you in advance for any help.
Morgane
Helpdesk (Haavard Rue)
do you have a copy of data as well that can be checked against ?
> You received this message because you are subscribed to the Google Groups "R-
> inla discussion group" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to r-inla-discussion...@googlegroups.com.
> To view this discussion, visit
> https://groups.google.com/d/msgid/r-inla-discussion-group/CACHvASYkKm4_K8mZWBiELB19sGcoOWSsAh2%3D%2BO%2BUQ%2B-v7rxaTw%40mail.gmail.com
> .
--
Håvard Rue
he...@r-inla.org
Finn Lindgren
This may just be a side-effect of a problem noticed by inla(), but in that case inlabru should be able to exit more cleanly.
--
You received this message because you are subscribed to the Google Groups "R-inla discussion group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to r-inla-discussion...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/r-inla-discussion-group/CACHvASYkKm4_K8mZWBiELB19sGcoOWSsAh2%3D%2BO%2BUQ%2B-v7rxaTw%40mail.gmail.com.