Dear Gretl list
Trying to improve an estimated model by OLS with a small sample, I tried to do a bootstrap on the coefficients and p-values after estimate by OLS model.
After reading the papers by John Fox "Bootstrapping Regression Models. Appenix to an R and S-Plus Companion to applied regressions" and Mc.Kinnon "Bootstrap methods in econometrics" I have seen that there are different kinds of bootstrap procedures (I never think that they could be so many bootstraps procedures).
I see in the gretl-GUI menu that after model estimation by OLS there is an option that allows bootstrap confidence intervals of the coefficients of the model, studentized intervals and p-value of the coefficients of the model.
Of the options available in this menu and the information in Gretl guide, it seems clear that this procedure is a "residual bootstrap" where the matrix of regressors X is treated as fixed. Is there any way by GUI option to do a "bootstrap case" where the regressors X are treated as "random"?. I have don this exporting data to R and using Boot functions from package "car".
One more question, how are treated missing values generated by lagged variables in the model in gretl bootrstrap procedure, are simply ignored?
And two last questions, do you think that this boostrap could be applicable in the context of time series analysis, which is my case, or would be better to use a blocking bootstrap? In this case, sombbody has script any code to do this?