Usually the bootstrap on time series is performed in the block manner:
you select blocks of consecutive observations long enough that the
correlation structure of the original series is mostly preserved, but
also that the observations on the ends of the block are approximately
independent. In your case, the blocks of length say 4 might be
reasonable. (Scholar.)Google on the block bootstrap.
I have an ma(1) garch(1,1) (we believe) original process.
We would like to generate probability distributions from the generated
equation and take random draws on the sample which would respect the original
process.
I believe using simple bootstrap on a time series cause the time structure of
the original series to be lost. But apparently, there is a way round it
(residuals resampling?)
--
Stas Kolenikov
http://stas.kolenikov.name
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
