Hi Spencer, you are quite right. I should have been careful to emphasize that the strategy I suggested was intended only to produce the "test for no correlation" clause of the "either a confidence interval or a test for no correlation" sentence.
Cheers Andrew On Mon, Sep 04, 2006 at 04:00:44PM -0700, Spencer Graves wrote: > Hi, Andrew: > > This will produce a "permutation distribution" for the correlation > under the null hypothesis of zero correlation between the variables. > This is a reasonable thing to do, and would probably produce limits more > accurate than the dashed red lines on the 'ccf' plot. However, they > would NOT be confidence interval(s). > > For a confidence interval on cross correlation, you'd have to > hypothesize some cross correlation pattern between x and y, preferably > parameterized parsimoniously, then somehow determine an appropriate > range of values consistent with the data. By the time you've done all > that, you've effectively fit some model and constructed confidence > intervals on the parameter(s). > > Best Wishes, > Spencer > > Andrew Robinson wrote: > >Jun, > > > >If your interest is to estimate the correlation and either a > >confidence interval or a test for no correlation, then you might try > >to proceed as follows. This is a Monte-Carlo significance test, and a > >useful strategy. > > > >1) use ccf() to compute the cross-correlation between x and y. > > > >2) repeat the following steps, say, 1000 times. > > > >2a) randomly reorder the values of one of the time series, say x. > > Call the randomly reordered series x'. > > > >2b) use ccf() to compute the cross-correlation between x' and y. > > Store that cross-correlation. > > > >3) the 1000 cross-correlation estimates computed in step 2 are all > > estimating cross-correlation 0, conditional on the data. A > > two-tailed test then is: if the cross-correlation computed in step > > 1 is outside the (0.025, 0.975) quantiles of the empirical > > distribution of the cross-correlations computed in step 2, then, > > reject the null hypothesis that x and y are uncorrelated, with size > > 0.05. > > > >I hope that this helps. > > > >Andrew > > > > > >Juni Joshi wrote: > > > >> Hi all, > >> > >> I have two time series data (say x and y). I am interested to > >> calculate the correlation between them and its confidence interval (or > >> to test no correlation). Function cor.test(x,y) does the test of no > >> correlation. But this test probably is wrong because of autocorrelated > >> data. > >> > >> ccf() calculates the correlation between two series data. But it does > >> not provide the confidence intervals of cross correlation. Is there > >> any function that calculates the confidence interval of correlation > >> between two time series data or performs the test of no correlation > >> between two time series data. > >> > >> Thanks. > >> > >> Jun > >>______________________________________________ > >>[email protected] mailing list > >>https://stat.ethz.ch/mailman/listinfo/r-help > >>PLEASE do read the posting guide > >>http://www.R-project.org/posting-guide.html > >>and provide commented, minimal, self-contained, reproducible code. > >> > >> > > > > -- Andrew Robinson Department of Mathematics and Statistics Tel: +61-3-8344-9763 University of Melbourne, VIC 3010 Australia Fax: +61-3-8344-4599 Email: [EMAIL PROTECTED] http://www.ms.unimelb.edu.au ______________________________________________ [email protected] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
