Every supplied value is treated as public protocol or schema metadata and is not charged to the privacy budget. Omitted regimen and timing quantities are estimated through budgeted private summaries.
Usage
pmx_public_design(
schema,
dose_times = NULL,
dose_interval = NULL,
n_doses = NULL,
dose_amount = NULL,
dose_rate = NULL,
infusion_duration = NULL,
dose_evid = 1,
dose_cmt = 1,
endpoint_grids = NULL,
endpoint_occasion_grids = NULL,
endpoint_cmt = NULL,
category_levels = NULL,
defaults = NULL,
time_jitter_sd = 0.02,
subject_count = NULL
)Arguments
- schema
A schema from
pmx_schema(). It is required.- dose_times, dose_interval, n_doses, dose_amount, dose_rate
Optional public regimen overrides. Normally omit these so the budgeted event summaries infer the regimen from the input data. Supply them only when they are independently public protocol facts, never by inspecting confidential records outside the private fit.
- infusion_duration
Optional public infusion duration.
- dose_evid, dose_cmt
Public event and dose-compartment values.
- endpoint_grids
Optional named list of fixed discretization bases on each endpoint's declared scientific clock. These are not sampling schedules. When omitted, the package constructs generic bases from public bounds and contribution limits and privately learns their occupancy.
- endpoint_occasion_grids
Optional named list of endpoint-specific, public dose-occasion sampling schedules. This exceptional override should be used only for an independently public protocol. Each endpoint entry is a list named by positive occasion number; omitted occasions generate no observations.
- endpoint_cmt
Named list or vector of public observation compartments.
- category_levels
Named lists of allowed values for categorical covariates and strata. Supplying levels forces even a numeric covariate to be treated categorically. Factor levels may instead come from the public schema. These levels are public domains, not values discovered by inspecting confidential records outside the private fit.
- defaults
Named values for otherwise unmodeled public columns.
- time_jitter_sd
Nonnegative generation-time jitter scale, expressed as a fraction of the closest public grid spacing.
- subject_count
Optional independently public source subject count. When omitted, generation uses the privacy-accounted fitted count.