Posterior predictive distribution of an edaphos Active-Learning fit

Samples the conditional distribution of the target at every row of candidates by asking the underlying Quantile Regression Forest (ranger) for a grid of quantiles. The returned edaphos_posterior carries those quantile samples directly, so uncertainty_calibrate() and ggplot2::autoplot() work without changes.

Usage

active_learning_posterior(model, newdata, n_quantiles = 99L, units = NULL)

Arguments

model

An edaphos_al_model produced by al_fit().

newdata

A data frame with (at least) the columns used as covariates at fit time.

n_quantiles

Integer; size of the equally-spaced grid of quantiles to request from the QRF. Defaults to 99L (1 % to 99 % in 1 % steps, which is a reasonable trade-off between a smooth empirical CDF and ranger::predict() cost).

units

Optional free-text units tag.

Value

An edaphos_posterior with method = "ensemble" (the QRF conditional distribution being itself an ensemble over tree leaves) and query_type = "sample".

References

Meinshausen, N. (2006). Quantile regression forests. Journal of Machine Learning Research 7, 983-999.