Yearly trend (d01 Poisson CIs) + Alert x Cause / Housing contingency chi^2 + Cramer's V on d06 / d07.
Examples
# \donttest{
otis_d01 <- morie_synth_otis("d01", n = 120L, seed = 1L)
otis_d06 <- morie_synth_otis("d06", n = 120L, seed = 1L)
otis_d07 <- morie_synth_otis("d07", n = 120L, seed = 1L)
morie_otis_analyze_d_chi2(list(d01 = otis_d01,
d06 = otis_d06, d07 = otis_d07))
#> OTIS d-series -- yearly death counts + Alert x Cause/Housing chi^2
#> ==================================================================
#> d01 total deaths 120
#> d01 year range 2018-2024
#> Year-over-year RR 1.000 (2024/2018; n=19/19; 95% CI [0.529, 1.889])
#> d06 (Alert x MedicalCause) Cramer's V 0.2226
#> d07 (Alert x Housing_Type) Cramer's V n/a
#>
#> d01 yearly custodial-death counts (Poisson 95% CI):
#> Year Deaths 95% CI
#> --------------------------
#> 2018 19 [11.4, 29.7]
#> 2019 18 [10.7, 28.4]
#> 2020 13 [6.9, 22.2]
#> 2021 15 [8.4, 24.7]
#> 2022 15 [8.4, 24.7]
#> 2023 21 [13.0, 32.1]
#> 2024 19 [11.4, 29.7]
#>
#> d06 / d07 Alert-flag x outcome contingency chi^2 + Cramer's V:
#> Dataset Slice n chi^2 dof p Cramer's V min cell
#> ----------------------------------------------------------------------------------
#> d06 Alert x MedicalCause 4939 489.428 6 1.59e-102 0.2226 103
#> d07 Alert x Housing_Type 0 n/a 0 n/a n/a 0
#>
#> d-series carries no per-individual alert columns -- the Ruhela alert-complexity dual is structurally impossible. Natural alternatives are: (1) yearly death-count trends, and (2) Cramer's V on d06 / d07 aggregate contingency tables.
# }