Package index • cmrdesign

Applied CMR rules

User-facing rules that estimate a confidence set from pilot data and return a recommended main-wave allocation plus a regret certificate.

binary_rectangle_corners() binary_rectangle_regret() cmr_two_arm_from_rectangle() cmr_binary_from_rectangle(): Two-arm CMR from a variance rectangle
cmr_multiarm(): Shared-control multi-arm CMR assignment
cmr_multiarm_from_rectangle(): Multi-arm CMR from a variance rectangle
cmr_multiple_outcomes(): Multiple-outcome CMR assignment
cmr_proxy() cmr_delayed_outcome(): Proxy or delayed-outcome CMR assignment
cmr_stratified(): Stratified CMR assignment
cmr_stratified_from_rectangle(): Stratified CMR from a variance rectangle
cmr_two_arm() cmr_binary(): Two-arm Conditional Minimax Regret assignment
cmr_unbounded(): Unbounded-outcome CMR assignment
cmr_unbounded_from_rectangle(): Unbounded-outcome CMR from a variance rectangle
pilot_plan() cmr_plan(): Pilot/main-wave planning summary

Screens from Appendix E of the accompanying paper (Yamin 2026) for planning the pilot/main-wave split before collecting pilot data.

activation_threshold_bounded() activation_threshold_bernoulli(): Pilot-size activation thresholds
pilot_viability_band(): Pilot viability band
break_even_pilot_share(): Break-even pilot share

Expert and audit helpers for the confidence sets used by CMR rules.

rectangle_bernoulli_binary() rectangle_binary() rectangle_two_arm() rectangle_bernoulli_two_arm(): Two-arm confidence rectangles
rectangle_bounded_two_arm() rectangle_bounded_binary(): Bounded two-arm confidence rectangle
rectangle_multiarm(): Multi-arm confidence rectangle
rectangle_multiple_outcomes(): Multiple-outcome confidence rectangle
rectangle_proxy() rectangle_delayed_outcome(): Proxy or delayed-outcome confidence rectangle
rectangle_stratified(): Stratified confidence rectangle
rectangle_unbounded(): Unbounded-outcome confidence rectangle
variance_bounds_bernoulli_exact(): Exact Bernoulli variance bounds
variance_bounds_maurer_pontil() variance_bounds_martinez_taboada_ramdas(): Bounded-outcome variance bounds
variance_bounds_unbounded_mom(): Unbounded-outcome median-of-means variance bounds
folded_binomial_pmf() folded_binomial_tails(): Folded-binomial distribution for Bernoulli sample variances
binary_rectangle_corners() binary_rectangle_regret() cmr_two_arm_from_rectangle() cmr_binary_from_rectangle(): Two-arm CMR from a variance rectangle

Baseline allocation rules and regret/objective helpers for audits, examples, and simulations.

assign_balance() assign_multiarm_balance() assign_stratified_balance() assign_feasible_neyman() assign_trimmed_neyman() assign_additive_regularized_neyman() assign_exponential_regularized_neyman(): Baseline and comparator assignment rules
multiarm_variance_objective() multiarm_oracle_variance() assign_multiarm_neyman() multiarm_regret() multiarm_rectangle_vertices(): Multi-arm variance objectives and Neyman allocation
stratified_variance_objective() stratified_oracle_variance() assign_stratified_neyman() stratified_regret() stratified_rectangle_vertices(): Stratified variance objectives and Neyman allocation
variance_objective() oracle_variance() assign_neyman() regret(): Two-arm variance objectives and Neyman allocation
realize_allocation() print(<cmr_allocation>): Convert CMR target shares to integer allocation counts

Small helpers for checking CMR certificates, coverage, boundary choices, and realized gains.

coverage_indicator() certificate_valid() boundary_indicator() boundary_rate() saving_vs_balance() share_of_oracle_gain(): Diagnostic metrics for CMR simulations

Print and summary methods for compact CMR result objects.

print(<cmr_two_arm>) print(<cmr_unbounded>) print(<cmr_proxy>) print(<cmr_multiple_outcomes>) print(<cmr_multiarm>) print(<cmr_stratified>) summary(<cmr_two_arm>) summary(<cmr_unbounded>) summary(<cmr_proxy>) summary(<cmr_multiple_outcomes>) summary(<cmr_multiarm>) summary(<cmr_stratified>) print(<summary.cmr_result>): Print and summarize CMR results

cmrdesign cmrdesign-package: cmrdesign: Conditional Minimax Regret design rules