Package index • synpmx

Generating data

The generation modes. synpmx_avatar() blends real subjects, synpmx_pca() fits a basis of their profiles, and synpmx_model() estimates a population model and simulates from it; the remaining three require a public structural model.

synpmx_avatar(): Synthesize a structurally faithful PMX dataset (AVATAR-style)
synpmx_pca(): Summarize a PMX dataset and generate a synthetic one from the summary
synpmx_pca_summarize(): Summarize a trial into the quantities a synthetic copy is built from
synpmx_pca_generate(): Generate a synthetic PMX dataset from a trial summary
synpmx_model(): Estimate a population model and generate a synthetic dataset from it
synpmx_model_estimate(): Estimate a population model from a trial
synpmx_model_generate(): Generate a synthetic PMX dataset from a fitted model
synpmx_prior(): Generate a dataset from public inputs only
synpmx_calibrated(): Generate a dataset from a privately calibrated structural model
synpmx_empirical(): Generate a dataset from a dense differentially private release
synpmx_generate(): Draw another dataset from a release already paid for

Reading a trial summary

What synpmx_pca_summarize() read out of the source data: its inventory, the dosing and visit models, and the components over time.

pca_report(): What the PCA fit read out of the source data
pca_features(): Every feature the components are built on
pca_scores(): The score model each arm is generated from
pca_dosing(): The planned dose schedule each arm was generated from
pca_dose_rates(): The dose-modification rates each arm was generated from
pca_visits(): The visit model each arm was generated from
pca_components(): Component loadings over time, and the variance each component explains
pca_component_effect(): What each component does to a profile, on the scale the data are reported in

Reading a fitted model

What synpmx_model_estimate() read out of the source data: the estimated half, the candidates it chose between, and the apparatus it summarized.

model_report(): What a fitted model carries
model_candidates(): The candidate models the selection was made from
model_parameters(): The estimated parameters

Declaring the data

Roles, endpoints, schema, and bounds that describe an event table.

pmx_roles(): Declare pharmacometric column roles
pmx_endpoint(): Declare endpoint scientific-clock behavior
pmx_schema(): Capture a schema asserted to be public
pmx_bounds(): Declare public numeric domains for private PMX fitting
pmx_generated_roles(): Roles for tables produced by .generate_structural()

Public model and design inputs

Data-independent inputs for the model-based modes. See the model and data elicitation articles for how to produce these without reading data.

pmx_structural_model(): Declare a public structural model
pmx_trial_design(): Declare a public trial design
pmx_public_design(): Declare public event-design information
pmx_prior(): Declare one public prior range
pmx_priors(): Collect public priors for the released corrections
pmx_covariate(): Declare one public baseline covariate
pmx_covariates(): Collect public covariate declarations
pmx_covariates_auto(): Declare bootstrap-resampled covariates by column name

Privacy accounting

Contribution limits, budget, preflight, and the release ledger.

pmx_contribution_limits(): Declare subject contribution limits
pmx_budget_allocation(): Allocate an epsilon budget across private summary groups
pmx_preflight(): Check whether a private release is worth its budget, before spending it
privacy_report(): Summarize a fitted model's privacy contract
validate_private_model(): Validate a fitted private PMX population model
dp_backend_status(): Inspect the differential-privacy backend
run_dp_backend_tests(): Run canonical checks against the configured DP backend
synpmx_enable_dp_engines(): Acknowledge the DP engines' unaudited status for this session
synpmx_disable_dp_engines(): Withdraw the acknowledgment from synpmx_enable_dp_engines()

Validation and diagnostics

Structural checks on generated data, and restricted comparisons.

synpmx_scorecard(): The scorecard for one synthetic dataset
synpmx_scorecard_datatable(): A scorecard as a coloured HTML table
validate_pmx(): Validate a pharmacometric event dataset
pmx_endpoint_types(): What kind of values each endpoint takes
compare_pmx(): Compare source and generated PMX structures inside the restricted environment
compare_pmx_distributions(): Compare per-covariate and per-endpoint distributions of source and synthetic
compare_pmx_distributions_height(): How tall the distribution figure should be drawn
compare_pmx_rare_levels(): Which rare source levels reached the synthetic output
compare_pmx_strata_sizes(): Stratum sizes, source against synthetic
compare_pmx_strata_endpoints(): Endpoints held by each stratum, source against synthetic
skeleton_uniqueness(): Score how many patients share each patient's event skeleton
plot_pmx_schedule(): Draw a cohort's dosing and observation schedule
unmaskable_strata(): Which strata can mask their own avatars
pmx_masking_report(): Report what each masking mechanism did, and what it cost
compare_pmx_proximity(): Are synthetic subjects sitting too close to real ones?
flag_identifiable_subjects(): Flag structurally unusual – and so easily identifiable – subjects
remediate_identifiable_subjects(): Remove or shorten the subjects flag_identifiable_subjects() flags
sampling_summary(): Summarize the fitted sampling design
strata_summary(): Summarize fitted strata and associated regimens

Fixtures

Fully public example datasets for testing and demonstration.

pmx_censoring_fixture(): Public PMX censoring fixture
pmx_simulated_fixture(): Fully simulated public repeated-dose fixture