Bootstrapping

Residual bootstrapping

Basically, you fit a model and resample the residuals with replacement from the original model. You then add the residuals back into the predicted outcome. Recompute again

Assumptions

1. The model errors are identically distributed (homoskedatic)- if you violate this than it’s inviting trouble
2. Relationship between predictor and dependent variable is linear

References

1. https://library.virginia.edu/data/articles/bootstrapping-residuals-linear-models-heteroskedastic-errors-invites-trouble#fn3