ABSTRACT
An alternative, known as the parametric bootstrap, is identical to the above except the disturbances are drawn from a particular distribution with its parameters set as estimates from the residuals, most commonly a normal random number generator with mean zero and variance equal to the residual variance. It is also possible to generate the disturbances as weighted averages of draws from the empirical distribution function of the residuals and the parametric distribution; this method may be called the smoothed bootstrap with the weights determining the degree of smoothing. Other techniques in Monte Carlo analysis, such as importance sampling and antithetic variates, can also be applied to the bootstrap. (See Efron and Tibshirani 1993 for some discussion and references.) These alternatives have the same essential properties as the ordinary bootstrap, which will remain our focus as it is the method of resampling that has almost exclusively been employed in econometric applications.
