Package index • proxymix

Top-level interface

The unified fitting verb and its dispatcher.

The S7 class hierarchy.

autoplot.gmm_fit: Plot a fitted Gaussian-mixture proxy
glance.gmm_fit: Glance at a fitted Gaussian-mixture proxy
gmm(): A Gaussian mixture
gmm_counterfactual_law(): A per-unit counterfactual law
gmm_dim(): Dimension of a Gaussian mixture
gmm_fit(): A fitted Gaussian-mixture proxy
gmm_n_components(): Number of components in a Gaussian mixture
gmm_target(): A target density on R^p
gmm_weights() gmm_means() gmm_covariances(): Component parameters of a Gaussian mixture
is_proposal(): An importance-sampling proposal
tidy.gmm: Tidy a Gaussian mixture into a component table

Ways to build a gmm_target — from samples, from a log-density, or from a built-in.

banana_target(): Banana-shaped 2-D target
donut_target(): Donut-shaped 2-D target
epanechnikov_target(): Compact-support Epanechnikov target
gmm_target_from_samples(): Build a target from samples alone
maxent_target(): Maximum-entropy target under moment and support constraints
mixture_target(): Three-component Gaussian-mixture target

The three KL-optimal fitting regimes from Hoek and Elliott (2024).

fit_em_samples(): Classical EM fit on samples
fit_kld_em(): Importance-sampled KLD-EM fit (regime iii)
fit_moment_match(): Closed-form moment-matching fit
from_kde(): Compile a kernel-density estimate into a Gaussian-mixture proxy
from_objective(): Map the optima of an objective with a Gaussian-mixture proxy
select_N(): Select the number of mixture components

Density, sampling, marginalisation, conditioning, divergence.

dgmm(): Density of a Gaussian mixture
gmm_canonicalise(): Canonicalise the component ordering of a Gaussian mixture
gmm_conditionalise(): Conditional of a Gaussian mixture
gmm_divergence(): Divergence between two Gaussian mixtures
gmm_kld(): Kullback-Leibler divergence between two Gaussian mixtures
gmm_marginalise(): Marginal of a Gaussian mixture
gmm_mean() gmm_cov(): Mean and covariance of a Gaussian mixture
gmm_modes(): Modes of a Gaussian mixture
pgmm() qgmm(): Distribution and quantile functions of a one-dimensional mixture
rgmm(): Sample from a Gaussian mixture

Pushforward, Kalman update, conditioning, the do-operator and the counterfactual.

gmm_affine(): Affine pushforward of a Gaussian mixture
gmm_aggregate(): Aggregation pushforward of a Gaussian mixture
gmm_convolve(): Convolution of two independent Gaussian mixtures
gmm_counterfactual(): Counterfactual law of one unit (abduction, action, prediction)
gmm_filter(): Bounded Gaussian-sum filtering over an observation series
gmm_intervene(): Interventional law of a Gaussian mixture (the do-operator)
gmm_missing(): Condition a Gaussian mixture on the exact values of some coordinates
gmm_mix(): Mix Gaussian mixtures into one mixture
gmm_observe(): Bayesian update of a Gaussian mixture on a noisy linear observation
gmm_product(): Pointwise product of two Gaussian mixtures
gmm_reduce(): Reduce a Gaussian mixture to fewer components

One joint fit read as CATE, optimal action, off-line policy value, and an identification audit.

fit_uplift(): Fit an uplift / next-best-action model from a data frame
gmm_cf_mean(): The identified counterfactual mean
gmm_cf_tail_prob(): Refused: a tail probability of an individual counterfactual law
gmm_cf_variance(): Refused: the variance of an individual counterfactual law
gmm_counterfactual(): Counterfactual law of one unit (abduction, action, prediction)
gmm_intervene(): Interventional law of a Gaussian mixture (the do-operator)
proxy_cate(): Heterogeneous treatment effects (CATE / uplift)
proxy_confounding_gap(): Confounding gap: the sensitivity of the effect to the latent regime
proxy_decide(): Optimal action and expected incremental value per unit
proxy_identification_report(): The identification report (an executive one-pager)
proxy_overlap(): Per-unit overlap / positivity diagnostic
proxy_policy_value(): Off-line value of a targeting policy
proxy_predict(): Predicted outcome under a treatment (the seeing rung)
proxy_regime_segments(): The fitted regimes as an interpretable segment table
proxy_retrospective_uplift(): Retrospective (counterfactual-mean) uplift for observed units
proxy_uplift(): Uplift (alias of proxy_cate() for a binary treatment)
uplift_identification(): Identification-report object
uplift_model(): A fitted uplift / next-best-action model

Multiple imputation by mixture conditioning, with pooling and the fraction of missing information.

as_mids(): Convert imputations to a mice multiply-imputed dataset
gmm_complete(): Extract completed datasets from a gmm_imputation
gmm_imputation(): A Gaussian-mixture multiple-imputation result
gmm_impute(): Multiple imputation by Gaussian-mixture conditioning
mar() mnar() censored(): Missingness mechanisms for multiple imputation
proxy_fmi(): Fraction of missing information for a column mean
proxy_mnar_sensitivity(): Missing-not-at-random sensitivity analysis for a coordinate mean
proxy_pool(): Pool a column mean across imputations

End-of-sample structural-break test on the Gaussian-sum filter.

gmm_eos_test(): End-of-sample instability test on a Gaussian state-space filter

Pluggable proposals for regime (iii).

is_mvn(): Multivariate-normal proposal
is_mvt(): Multivariate-t proposal
is_uniform(): Uniform-on-a-box proposal
proposal_uniform() proposal_mvn() proposal_mvt(): Preferred names for the importance-proposal constructors

Compiling external Bayesian posteriors into proxies.

gmm_target_from_posterior(): Compile an unnormalised Bayesian posterior into a gmm_target

Initial guesses + best-of-multistart wrapper.

KLD trace, ESS, Hellinger MC, information criteria.

bic_aic(): Information criteria: BIC, AIC, and ICL
ess_summary(): Summary of importance-sampling diagnostics
ess_trace(): Effective sample size of the importance-sampling weights
gmm_anneal_path(): Phase-transition component discovery by deterministic annealing
gmm_conditional_entropy(): Conditional predictive entropy of a Gaussian mixture
gmm_entropy(): Differential entropy of a Gaussian mixture
gmm_evidence(): Estimate the target's normalising constant from a fitted proxy
gmm_fit_ensemble(): Bootstrap ensemble of a fitted proxy
gmm_fit_quality(): The quality certificate of a fit or derived mixture
gmm_independence_graph(): Conditional-independence (Gaussian graphical model) structure of a mixture
gmm_mutual_information(): Cauchy-Schwarz mutual information between two coordinate blocks
hellinger_mc(): Monte-Carlo Hellinger distance between a fit and its target
kld_trace(): Per-iteration KLD trace of a fit
proxy_functional_ci(): Percentile interval for any functional of a fitted proxy