10
$\begingroup$

I currently use the following process for bootstrapping a multivariate time series in R:

  1. Determine block sizes -- run the function b.star in the np package which produces a block size for each series
  2. Select maximum block size
  3. Run tsboot on any series using the selected block size
  4. Use index from bootstrap output to reconstruct multivariate time series

Someone suggested using the meboot package as an alternative to the block bootstrap but since I am not using the entire data set to select a block size, I am unsure of how to preserve correlations between series if I were to use the index created by running meboot on one series. If anyone has experience with meboot in a multivariate setting, I would greatly appreciate advice on the process.

$\endgroup$
7
$\begingroup$

First, I want to highly recommend the maximum entropy bootstrap (meboot). I abandoned the block bootstrap in favor of meboot, and I've been very pleased with the results. The algorithm does not use blocking in any way, it does not require stationarity, and yet it incorporates the correlation structure of the data. It's cool.

Second, while I confess that I've never done a multivariate bootstrap using meboot, I believe you can recast your time series data as panel data and use the meboot.pdata.frame function to perform an essentially multivariate bootstrap.

$\endgroup$
  • $\begingroup$ I think the function is meboot.pdata.frame. $\endgroup$ – ProbablePattern Aug 5 '11 at 11:43
  • $\begingroup$ Yikes! Thanks, P.P. Yes, I misspelled the function, and the correct name is meboot.pdata.frame. Sorry 'bout that. $\endgroup$ – pteetor Aug 6 '11 at 16:28
  • $\begingroup$ I came across this thread and got interested in the ME bootstrap. Playing with it a bit, I found that it replicated time series quite closely -- not a lot of variability in the bootstrap DGP. Digging further, I found the following paper by Davidson that includes a critique of the ME Bootstrap: www.monticini.eu/wp/rdavidson.pdf. Not sure if there has been any rebuttal of it, but given my own simulation results, I'd be hesitant to use it. $\endgroup$ – generic_user Apr 29 '16 at 22:47
  • $\begingroup$ @generic_user Thank you for that reference! I read it with great interest. I, too, have noted the lack of variability in the bootstrap replicates under certain circumstances. I'm not yet convinced that, as a result, the ME bootstrap is never valid. But the paper is sobering. More investigation is needed. $\endgroup$ – pteetor May 5 '16 at 11:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.