Perturbation family: delta-adjusted multiple imputation (tipping point)

Imputes the missing responses by fully conditional specification (regression method, one column per visit, subject-level covariates and treatment as predictors), adds delta to every imputed value of the active arm, refits the pre-specified model on each completed data set and pools with Rubin's rules. delta = 0 is the missing-at-random multiple-imputation analysis and is a counted alternative; other deltas are tipping-point searches (kind = "search"). Rows carry the delta in param.

Usage

stress_tipping(
  spec,
  deltas = seq(0, -10, by = -2),
  m = 25,
  seed = 1,
  arm = NULL,
  shift_visits = NULL,
  impute_covariates = NULL,
  nbiter = 10,
  shift_every = TRUE,
  cores = 1L
)

Arguments

spec: A stress_spec().
deltas: Shifts added to the imputed values of the active arm, in response units.
m: Number of imputations.
seed: Random seed (the same seed is used for every delta). The m imputations are drawn in turn from one stream started at seed, so different seeds give unrelated imputations.
arm: Treatment level receiving the shift; defaults to the active arm.
shift_visits: Visit levels whose imputed values are shifted; NULL shifts every visit.
impute_covariates: Subject-level covariates used as predictors in the imputation model (with treatment and the visit columns); NULL uses every spec covariate that is constant within subject. Covariates that vary within subject cannot be used.
nbiter: Burn-in iterations of the FCS sampler.
shift_every: If TRUE the shift is applied at every draw (SAS MNAR ADJUST under FCS); if FALSE only to the final imputation.
cores: Number of worker processes; 1 fits sequentially.

Value

A stress_rows data frame, one row per delta; note records the number of imputations and the Monte Carlo standard error of the pooled estimate.

Details

Covariates that vary within subject are excluded from the imputation model and are kept in the analysis model; subject-visit rows absent from the data are created for the imputed response with those covariates missing, so they drop out of the refit.

Examples

small <- mmrm::fev_data[mmrm::fev_data$USUBJID %in% paste0("PT", 1:80), ]
spec <- stress_spec(small, "FEV1", "USUBJID", "AVISIT", "ARMCD", "PBO", covariates = ~ FEV1_BL)
stress_tipping(spec, deltas = c(0, -5), m = 3)
#>                               family    label        kind        role is_base
#> 1 delta-adjusted multiple imputation  delta 0 alternative sensitivity   FALSE
#> 2 delta-adjusted multiple imputation delta -5      search sensitivity   FALSE
#>   converged
#> 1      TRUE
#> 2      TRUE
#>                                                                                note
#> 1 shift on TRT at all visits; seed 1; m = 3 of 3 converged; MC SE of estimate 0.681
#> 2 shift on TRT at all visits; seed 1; m = 3 of 3 converged; MC SE of estimate 0.603
#>   estimate   stderr       df       lcl      ucl         p n_subj n_obs aicc
#> 1 3.885759 2.670879 29.54786 -1.572405 9.343924 0.1562495     80   320   NA
#> 2 2.532657 2.658249 47.06471 -2.814856 7.880170 0.3455825     80   320   NA
#>   param covariance        method                 vcov covariates alpha
#> 1     0         us Kenward-Roger Kenward-Roger-Linear    FEV1_BL  0.05
#> 2    -5         us Kenward-Roger Kenward-Roger-Linear    FEV1_BL  0.05