Synthetic datasets

Generators return data together with known population quantities and are intended for examples, tests, and method studies.

Each generator accepts a sample size, random seed, and dataframe backend. Its return value contains the generated frame and a mapping of known population quantities.

Three functions serve the navigation program that the examples share.

function	what it returns
navigation_data	the make_nonlinear_bounded draw, with the columns under the program’s names
navigation_protocol	the StudyProtocol of the one-decision program
longitudinal_navigation_protocol	the StudyProtocol of the two-decision program

Eight names serve the subscription-retailer program of the offers tutorial.

name	what it holds
make_sequential_offers	a long panel of three weekly offers, with the logged probabilities and the exact truth. policy= draws a deployment-check cohort
offers_truth	the exact week-4 share under each plan, and each contrast with no offers
offers_protocol	the StudyProtocol of the campaign
win_back, win_back_then_ship	the two dynamic rules, as functions of the week’s engagement
OFFER_PLANS, OFFER_LEVELS, OFFER_WEEKS	the five plans by name, the three offer levels, and the number of weekly sends

make_sequential_offers returns a long panel and takes the keyword policy. It is not in GENERATORS, which holds generators of one wide frame with the columns Y and A.

Seven names serve the pricing tutorial of the same program.

name	what it holds
make_sequential_pricing	a wide frame of three monthly renewal prices with engagement, the renewal, and the revenue, with the exact truth
pricing_truth	the exact mean renewal and revenue under each plan, each effect against the engine’s own prices, and the decomposition of the effect by path
pricing_protocol	the StudyProtocol of the pricing question
price_rise	a plan that raises each price by a factor and keeps the engine’s price where the raised price would leave the band
PRICE_PLANS, PRICE_BANDS, PRICE_NODES	the three plans by name, the price band of each month, and the number of monthly prices

make_sequential_pricing is not in GENERATORS either. Its columns are usage, engaged_1, price_1, engaged_2, price_2, engaged_3, price_3, renewed, and revenue.

cleverly.datasets.available	Names of the bundled data-generating processes.
cleverly.datasets.make_linear_ate	Generate point-treatment data with a constant additive effect.
cleverly.datasets.make_nonlinear_ate	Nonlinear nuisance functions and a heterogeneous effect.
cleverly.datasets.make_nonlinear_bounded	Generate a bounded outcome with nonlinear nuisance functions.
cleverly.datasets.make_binary_outcome	A binary outcome with known risks, risk ratio and odds ratio.
cleverly.datasets.make_multi_arm	A three-armed confounded process with known counterfactual means.
cleverly.datasets.make_heterogeneous	A sample whose treated and control populations have genuinely different effects.
cleverly.datasets.make_instrument	Instrument-inflation: a covariate that predicts treatment but not the outcome.
cleverly.datasets.make_shift_dose	A confounded continuous dose with known means under each declared shift.
cleverly.datasets.make_missing_outcome	Missing outcomes, with a missingness mechanism that depends on treatment.
cleverly.datasets.make_missing_outcome_binary	A binary outcome with outcomes missing at random given (A, W).
cleverly.datasets.make_missing_outcome_bounded	Missing outcomes on the known support (0, 1), with a nonlinear outcome mean.
cleverly.datasets.make_cde	Controlled direct effect: the effect of A with Z held fixed.
cleverly.datasets.make_clustered	Clustered data with an unobserved shared effect modifying the treatment effect.
cleverly.datasets.make_biased_sample	A sample selected with known, unequal probabilities from a known population.
cleverly.datasets.make_weak_overlap	A practical positivity violation.
cleverly.datasets.make_longitudinal	Two time points, a binary outcome and (optionally) monotone censoring.
cleverly.datasets.make_longitudinal_survival	Two time points, an absorbing event at each, and monotone censoring.
cleverly.datasets.make_longitudinal_repeated	Three time points, a repeated binary outcome at each, and monotone censoring.
cleverly.datasets.make_longitudinal_competing	Two time points, two competing absorbing causes at each, monotone censoring.
cleverly.datasets.make_longitudinal_weighted	A biased sample of make_longitudinal(), with the design weights that undo it.
cleverly.datasets.make_point_survival	A baseline treatment, a discrete-time event, and dropout at the visits.
cleverly.datasets.point_survival_truth	The exact truth of make_point_survival() under an intervention on \(A\).
cleverly.datasets.navigation_data	Draw the point-treatment program data, with the columns under the program's names.
cleverly.datasets.navigation_protocol	Return the protocol for the standard navigation offer in the tutorial program.
cleverly.datasets.longitudinal_navigation_protocol	Return the protocol for navigation offered at two decisions after discharge.
cleverly.datasets.make_sequential_offers	Draw a weekly-offer campaign as a long panel, with the exact truth of its plans.
cleverly.datasets.offers_truth	Return the exact week-4 full-price share under each plan, and each contrast with no offers.
cleverly.datasets.offers_protocol	Return the protocol of the weekly-offer campaign.
cleverly.datasets.win_back	Return the win-back offer: the discount for a customer who is not engaged, else nothing.
cleverly.datasets.win_back_then_ship	Return the win-back-then-ship offer: the discount if not engaged, else free shipping.
cleverly.datasets.make_sequential_pricing	Draw three monthly renewal prices and the renewal, with the exact truth of the plans.
cleverly.datasets.pricing_truth	Return the exact mean renewal and revenue under each plan, and the effect of each plan.
cleverly.datasets.pricing_protocol	Return the protocol of the monthly renewal-price program.
cleverly.datasets.price_rise	Return the plan that raises each price by factor and keeps the price in the band.