Package index • mixqr

Model fitting

The main entry point and its control settings. mixqr() also hosts the expectile and M-quantile component-loss families via its family = argument.

mixqr(): Fit a finite mixture of quantile regressions
mixqr_control(): Control parameters for mixqr()
mixqr_vcontrol(): Control parameters for stochastic-EM variance estimation (Algorithm 3.1)

Component-specific penalized selection of covariates.

mixqr_pen(): Fit a penalized (variable-selecting) mixture of quantile regressions
selectedVars(): Active (selected) covariates per component of a penalized mixqr fit

Joint estimation at several quantile levels with shared classification and guaranteed non-crossing.

mixqr_nc(): Fit a joint multi-tau mixture of quantile regressions (non-crossing, shared labels)
predict(<mixqr_multitau>): Predict non-crossing component quantiles from a multi-tau fit
plot(<mixqr_multitau>): Plot the non-crossing component quantile curves of a multi-tau fit
sim_mixqr_cross(): Simulate a two-component mixture whose per-tau fits cross (for O4 validation)

Extract coefficients, standard errors, intervals, and predictions.

summary(<mixqr>): Summarize a mixqr fit
coef(<mixqr>): Component coefficients of a mixqr fit
confint(<mixqr>): Confidence intervals for a mixqr fit
vcov(<mixqr>): Variance-covariance matrix of a mixqr fit
predict(<mixqr>): Predict from a mixqr fit
fitted(<mixqr>) residuals(<mixqr>) nobs(<mixqr>): Fitted values, residuals and number of observations
logLik(<mixqr>) AIC(<mixqr>) BIC(<mixqr>): Log-likelihood, AIC and BIC of a mixqr fit
plot(<mixqr>): Plot a mixqr fit

Reproducible designs and the bundled example dataset.

The engine contract and reusable building blocks for custom EM engines and companion packages (QMM sub-projects 03-06).

register_mixqr_engine(): Register a mixqr EM engine
get_mixqr_engine(): Retrieve a registered mixqr engine constructor
list_mixqr_engines(): List registered mixqr engines
weighted_rq(): Weighted quantile regression for one mixture component
constrained_kde(): Constrained KDE update for all components (eq. 2.5 unequal / eq. 2.8 equal).

mixqr-package: mixqr: An Extensible Framework for Mixtures of Quantile and Expectile Regressions