Cross-Validation,Jackknife, and BootstrapMethods for Obtaining Standard Errors

Another alternative to the propagation of error formulas is the jackknife method. It can provide a standard error estimate for almost any estimator. Both the jackknife and cross-validation procedures use estimates constructed from data sets that use all but one observation. Each observation is deleted once. Unlike cross-validation that is used to estimate prediction mean squared error, the jackknife provides standard errors for parameter estimates. e dierence is akin to the distinction between prediction intervals and condence intervals for regression at a xed value of x (refer to Chapter 10 for a full discussion).