Pipeline stage: assemble and simulate a multistage test

Builds a two-stage multistage test (MST) from the calibrated items: a routing module of medium difficulty, then n_modules second-stage modules targeted at increasing ability levels. Each module is assembled with eatATA to be most informative at its target. Simulees are routed to the second-stage module whose target is closest to their ability estimate after the routing module; their final ability is estimated from both modules.

Usage

assemble_mst(
  p,
  router_length = 10,
  module_length = 10,
  n_modules = 3,
  targets = NULL,
  n = 1000,
  solver = c("GLPK", "lpSolve"),
  time_limit = 20,
  seed = NULL
)

Arguments

p: a calibrated pipeline from calibrate().
router_length, module_length: number of items in the routing module and in each second-stage module.
n_modules: number of second-stage modules (2 or 3 are typical).
targets: ability targets of the second-stage modules. Defaults to evenly spaced standard normal quantiles.
n: number of simulees.
solver, time_limit: passed to eatATA's solver.
seed: optional random seed.

Value

The pipeline with p$forms (a list with method = "mst", modules, a tibble of module and item, targets, cuts (the routing cut points), persons and summary).

Details

The checks (thresholds from mv_rules()):

no_solution (block): the modules could not be assembled.
mst_routing (review): more than mst_routing of the simulees were routed to a different module than their true ability calls for.
cat_precision (review): the average final standard error is above cat_sem, or the average bias is beyond cat_bias.

Examples

# \donttest{
if (requireNamespace("eatATA", quietly = TRUE) &&
    requireNamespace("mirtCAT", quietly = TRUE)) {
  p <- add_responses(mv_pipeline(), certexam, items = 3:34, key = certexam_key)
  p <- calibrate(drop_items(p, "Q14"), model = "2PL")
  p <- assemble_mst(p, router_length = 8, module_length = 7, n_modules = 3,
                    n = 300, seed = 1)
  p$forms$summary
}
#> Warning: Forms is running with 1 review open flag(s) from earlier stages:
#>   #2 [review] local_dependence: Q21 & Q22 (Q3 = 0.27)
#> Flags only warn because strict = FALSE. See flags(p).
#> $n
#> [1] 300
#> 
#> $mean_se
#> [1] 0.4848705
#> 
#> $bias
#> [1] -0.0147774
#> 
#> $rmse
#> [1] 0.4636104
#> 
#> $misrouted
#> [1] 0.3166667
#> 
#> $routed
#> [1] 0.3166667 0.3033333 0.3800000
#> 
# }