Intro to Workboots: Make Prediction Intervals from Tidymodel Workflows
summary
workboots is an R package that turns any tidymodel workflow (including XGBoost) into interval predictions by fitting many bootstrap resamples, predicting on new data, and returning median, lower and upper bounds; example uses Palmer penguins and shows 2000 estimations per observation, default functions predict_boots and vi_boots produce prediction or confidence intervals and variable importance intervals, requires roughly 5–20 minutes to run and assumes roughly normal residuals, and the outputs summarize uncertainty across models rather than a single point estimate.
AI-written summary · about
chapters (4)
related talks
16mIntro to Workboots: Make Prediction Intervals from Tidymodel Workflows
22mConformal Inference with Tidymodels
27mI have no clue what my model is doing...
21mTidypredict with recipes, turn workflow to SQL, spark, duckdb and beyond
19mUnderstanding, Generating, and Evaluating Prediction Intervals
28mIID Got You Down? Resample Time Series Like A Pro
37mCreating correct and capable classifiers
45mUncertainty estimation at scale with functime