[R] how to deal with NaN in "boot" object

[Prof Brian Ripley]

On Fri, 5 Sep 2008, Jinsong Zhao wrote:

> Hi there,
>
> I use boot() to do bootstrap simulation on my statistic, I get a boot
> object, named "obj". For certain reasons, there are some "NaN" in obj$t.
> Now, I hope to get confidence interval using boot.ci(), it give the
> following error:
>
> Error in if (const(t, min(1e-08, mean(t)/1e+06))) { :
>  missing value where TRUE/FALSE needed
>
> Now, what I could do is I have to edit the obj by hand.
>
> Any suggestions? Thanks in advanced!

Correct your 'statistic' function not to return NaN.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595