Tests that the ELPD ranking is invariant to different FLOOR penalty values in the Stan model.
smoketest_floor_elpd_invariance(
DT,
stan_code,
floors = c(-1e+06, -1e+08, -10000),
spec = "C",
controls = character(0),
k_grid = 0:1,
hs_grid = data.frame(hs_tau0 = c(0.1, 0.5), hs_slab_scale = c(1, 5), hs_slab_df = 4),
hs_rows = 1:2,
iter_warmup = 200,
iter_sampling = 200,
chains = 2,
seed = 123,
verbose = TRUE
)A list with components:
Logical; TRUE if ranking is identical across all FLOOR values.
The tested FLOOR values.
List of result tables for each FLOOR.
Combined data.frame of all results.
Character vector of ranking signatures.
Data.table with the data.
Character; Stan model code.
Numeric vector of FLOOR values to test.
Character; model specification.
Character vector of control variables.
Integer vector of lag values to test.
Data.frame with horseshoe hyperparameter grid.
Integer vector; which rows of hs_grid to use.
Integer; warmup iterations.
Integer; sampling iterations.
Integer; number of chains.
Integer; random seed.
Logical; print progress messages.