Skip to contents

Pipeline stage: bank monitoring

Source: R/pipeline-monitor.R
monitor_bank.Rd

Keeps an item bank healthy from one administration to the next. Each run records the current administration in the bank (how many test takers saw each item, and the test's marginal reliability) under a window name, then checks the bank's history. Items it flags can be retired or sent back to review with retire_items(). Drift and compromise checks are in check_security().

Usage

monitor_bank(
  p,
  bank = NULL,
  window = p$title,
  date = Sys.Date(),
  content = NULL,
  min_items = NULL,
  record = TRUE
)

Arguments

p

a pipeline.

bank

the item bank, from mv_bank() or bank_items(); defaults to p$bank.

window

a name for the current administration, such as "2026 spring". Defaults to the pipeline title. Recording a window again replaces it. Use record = FALSE to check the bank without recording.

date

the date of the administration.

content

optional content cell of each bank item: a named vector (item id = cell). It is stored in the bank. By default the cells come from the bank, or from the items in the pipeline (add_items(), generate_items()).

min_items

optional fewest active items per cell: one number, or a named vector by cell.

record

logical; record the current administration in the bank. It needs a calibrated pipeline.

Value

The pipeline with the updated bank in p$bank and p$monitoring: windows (one row per administration), items (use of each bank item), and cells (active items per content cell). Save the bank with saveRDS(p$bank, ...) to keep its history.

Details

The checks (thresholds from mv_rules()):

  • over_exposure (review): active items seen by more than max_exposure of the test takers in each of the last exposure_windows windows.

  • bank_depth (review): a content cell with fewer active items than its minimum: min_items, or by default cell_depth times the cell's count in the blueprint from set_blueprint().

  • reliability_trend (review): marginal reliability in the latest window more than reliability_drop below its value in the first window.

  • unused_items (info): more than unused_share of the active items (that have been in the bank for unused_days) not used in the last unused_days days.

See also

retire_items()

Examples

# \donttest{
p <- add_responses(mv_pipeline("Spring"), certexam, items = 3:34, key = certexam_key)
p <- calibrate(drop_items(p, "Q14"), model = "2PL")
p <- bank_items(p, name = "Certification bank")
p <- monitor_bank(p, window = "2026 spring", date = as.Date("2026-04-15"))
#> Warning: Bank monitoring 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).
p$monitoring$windows
#> # A tibble: 1 × 4
#>   window      date           n reliability
#>   <chr>       <date>     <int>       <dbl>
#> 1 2026 spring 2026-04-15  1200       0.831
flags(p, stage = "monitoring")
#> # A tibble: 0 × 10
#> # ℹ 10 variables: id <int>, stage <chr>, check <chr>, target <chr>,
#> #   value <chr>, rule <chr>, level <chr>, advice <chr>, status <chr>,
#> #   note <chr>
# }