0

I'm trying to build bootstrapped confidence intervals for a correlation coefficient between two non-stationary time series in R. I'm currently using the moving blocks bootstrapping method from the tsboot package, but I read that it is actually not that well-suited for non-stationary time-series. Is there to handle this in the tsboot package, or some other package? I read this paper: https://www.sciencedirect.com/science/article/pii/S1631073X02025785

Here they deal with it by selecting only blocks that are ‘near’x in the original series, but I cannot find a application of this in the boot package in R and have no clue how to code this myself. So if someone has any suggestion how to handle this, thanks in advance!

DanielBulgar
  • 1
  • I am curios to know what type of correlation coefficient are you estimating between two time series. – kangaroo_cliff Dec 02 '20 at 00:16
  • Pearson, but maybe a bit more info is needed: I am actually trying to kind of replicate this paper: https://www.aeaweb.org/articles?id=10.1257/mac.20180005. Where they first calculate ten year moving averages for two time series and then calculate the pearson correlation coefficient between these two series. Finally, they provide confidence intervals for the coefficient which they estimate manually, which is a bit too advanced for me. Therefore, I'm trying to bootstrap this confidence intervals... – DanielBulgar Dec 02 '20 at 10:26

0 Answers0