Repository files navigation

conformalbayes provides functions to construct finite-samplecalibrated predictive intervals for Bayesian models, following theapproach inBarber et al. (2021).These intervals are calculated efficiently using importance sampling forthe leave-one-out residuals. By default, the intervals will also reflectthe relative uncertainty in the Bayesian model, using thelocally-weighted conformal methods ofLei etal. (2018).

You can install the development version ofconformalbayes with:

# install.packages("devtools")devtools::install_github("CoryMcCartan/conformalbayes")

library(rstanarm)library(conformalbayes)data("Loblolly")fit_idx= sample(nrow(Loblolly),50)d_fit=Loblolly[fit_idx, ]d_test=Loblolly[-fit_idx, ]# fit a simple linear regressionm= stan_glm(height~ sqrt(age),data=d_fit,chains=1,control=list(adapt_delta=0.999),refresh=0)# prepare conformal predictionsm= loo_conformal(m)# make predictive intervalspred_ci= predictive_interval(m,newdata=d_test,prob=0.9)print(head(pred_ci))#>             5%       95%#> 1  -0.15888598  5.600095#> 29 25.43314599 30.988491#> 57 48.67648127 54.182655#> 2  -0.09561987  5.447242#> 30 25.42970114 30.938488#> 72 58.01173187 63.596592# are we covering?mean(pred_ci[,"5%"]<=d_test$height&d_test$height<=pred_ci[,"95%"])#> [1] 0.9117647

Read more on theGetting Startedpage.

Barber, R. F., Candes, E. J., Ramdas, A., & Tibshirani, R. J. (2021).Predictive inference with the jackknife+.The Annals of Statistics,49(1), 486-507.

Lei, J., G’Sell, M., Rinaldo, A., Tibshirani, R. J., & Wasserman, L.(2018). Distribution-free predictive inference for regression.Journalof the American Statistical Association, 113(523), 1094-1111.