bartz.testing

Testing utilities to generate synthetic datasets.

Data generation

gen_data(key, *, n, p[, k, lambda_, offset, ...])	Generate data from a quadratic multivariate DGP.
gen_params(key, *, p, k, q[, lambda_, ...])	Sample DGP coefficients and parameters (no dependence on n).
gen_data_from_params(key, params, *, n)	Sample predictors and outcomes given fixed params.
Params(partition, beta_shared, ...[, het_shape])	All quantities of the data-generating process that do not depend on n.
DGP(x, y, z, mulin_shared, mulin_separate, ...)	Output of gen_data / gen_data_from_params: sampled data and parameters.
QuantizedData(x, y, max_split)	Output of DGP.quantize: data in the format of bartz.mcmcstep.init.

Distributions

Standardized distributions (mean 0, variance 1) for predictors and coefficients.

Distr()	Family of standardized distributions: mean 0, variance 1.
Normal()	Standard Normal distribution.
Uniform()	Continuous uniform distribution, standardized: U(-sqrt(3), sqrt(3)).
DiscreteUniform(m)	Uniform distribution on m equispaced levels, standardized.

Scale distributions

Distributions of the per-predictor importance scales (unit mean square).

ScaleDistr()	Family of nonnegative scale distributions, normalized to E[s ** 2] = 1.
Constant()	Scales concentrated at 1 (uniform predictor importance).
Gamma(alpha)	Gamma(alpha) scales, rescaled to E[s ** 2] = 1.
SpikeSlab(pi)	Two-point distribution over the scales 0 and 1/sqrt(pi).