Moving quantile()?

On Thu, 26 Nov 2009, Rob Steele wrote:

Is there a faster way to get moving quantiles from a time series than to
run quantile() at each step in the series?

Yes.

Run

    help.request()

Since you have already done the first 4 items below (right?) you will
answer 'y', but when you get to the fifth one (which you evidently have
not done), you will have to answer 'n'.

----

Checklist:
Have you read the posting guide? (y/n) y
Have you checked the FAQ? (y/n) y
Have you checked An Introduction to R? (y/n) y
Have you checked the NEWS of the latest development release? (y/n) y
Have you looked on RSiteSearch? (y/n) n

----

This will start a browser with the RSiteSearch URL loaded.

Now comes the tricky part --- you have to type something into the
'Query' box and then click on 'Search'. Perhaps

    moving quantiles

will work??

It did for me, and the function I found by that strategy seems to be
faster than

    apply(embed(x[1:10000],100),1,quantile,.75)

by about 3 orders of magnitude.

HTH,

Chuck