Week-over-week QC: settled-data integrity (A) + frontier status signal (B)
Source:R/compare.R
qc_week_over_week_v1.RdWeek-over-week QC: settled-data integrity (A) + frontier status signal (B)
Usage
qc_week_over_week_v1(
current,
previous,
max_delay,
tol = 1e-06,
status_roles = c("status", "hlmstatus")
)Arguments
- current, previous
Two runs' collapsed csfmt.
- max_delay
Nowcast horizon (weeks); sets the settled/frontier boundary.
- tol
Tolerance for "unchanged" in the integrity check.
- status_roles
Naming-grammar roles treated as ORDINAL STATUS rather than as continuous medians: excluded from `$integrity`, and the only roles whose transitions appear in `$signal`. Defaults to both status-writing roles in the package – `"status"` from [mem_thresholds_v1] and `"hlmstatus"` from [signal_detection_hlm]. Before this argument existed, only `"status"` was selected. HLM alert transitions were then silently dropped from `$signal`, and HLM status columns were wrongly diffed as continuous values in `$integrity`.
See also
Neither package vignette covers run-over-run comparison.
compare_results is the underlying diff, and documents which
identity column names this check needs.
qc_surveillance_data_v1 answers a different question, about one
input feed rather than two finished runs.
Examples
w <- cstime::dates_by_isoyearweek$isoyearweek
i <- match("2023-01", w)
set.seed(1)
d <- data.table::data.table(
isoyearweek_reference = w[i + rep(0:39, each = 3)],
isoyearweek_reporting = w[i + rep(0:39, each = 3) + rep(0:2, 40)],
numerator = rpois(120, c(30, 15, 5)),
indicator_tag = "x", location_code = "nation", age = "total", sex = "total"
)
d <- d[isoyearweek_reporting <= w[i + 39]]
id <- c("indicator_tag", "location_code", "age", "sex")
# seed inside `run()`, so the two runs differ only by the data revision and
# not by their Monte-Carlo draws
run <- function(x) {
set.seed(2)
ens_collapse(nowcast_quasipoisson_v1(
csfmt_reporting_triangle_v3(x, id_cols = id),
max_delay = 3, n_sim = 200
))
}
cur <- run(d)
d_prv <- data.table::copy(d)
d_prv[isoyearweek_reference == w[i + 10], numerator := numerator - 5]
#> Index: <isoyearweek_reference>
#> isoyearweek_reference isoyearweek_reporting numerator indicator_tag
#> <char> <char> <int> <char>
#> 1: 2023-01 2023-01 26 x
#> 2: 2023-01 2023-02 20 x
#> 3: 2023-01 2023-03 8 x
#> 4: 2023-02 2023-02 38 x
#> 5: 2023-02 2023-03 16 x
#> ---
#> 113: 2023-38 2023-39 21 x
#> 114: 2023-38 2023-40 3 x
#> 115: 2023-39 2023-39 24 x
#> 116: 2023-39 2023-40 20 x
#> 117: 2023-40 2023-40 20 x
#> location_code age sex
#> <char> <char> <char>
#> 1: nation total total
#> 2: nation total total
#> 3: nation total total
#> 4: nation total total
#> 5: nation total total
#> ---
#> 113: nation total total
#> 114: nation total total
#> 115: nation total total
#> 116: nation total total
#> 117: nation total total
prv <- run(d_prv)
qc <- qc_week_over_week_v1(cur, prv, max_delay = 3)
# A settled week whose published median moved between runs. This table is
# ideally empty; a row in it means history was rewritten.
qc$integrity
#> indicator_tag isoyearweek column prv cur abs_diff
#> <char> <char> <char> <num> <num> <num>
#> 1: x 2023-11 numerator_nowcasted_q50x0 31 46 15
# Status transitions on the frontier weeks. Empty here because these runs
# carry no status column at all: mem_thresholds_v1() writes role "status" and
# signal_detection_hlm() writes role "hlmstatus", and neither was run. Both
# roles are selected by default -- see `status_roles`.
qc$signal
#> Empty data.table (0 rows and 5 cols): indicator_tag,isoyearweek,column,prv,cur