R help - Feb 2012 - smooth.spline() unique 'x' values error

Hello. 

I'm getting an unexpected result when running smooth.spline().  Here is a
simple example that replicates the error I'm getting:
> aa <- c(1, 2, 3, 8, 8, 8, 8, 8, 8, 8, 8, 8, 12, 13, 14)
> bb <- 1:length(aa)
> plot(aa, bb)
> smooth.spline(aa, bb)
Error in smooth.spline(aa, bb) : need at least four unique 'x' values



As you can see from the example, my 'x' clearly has more than 4 values. 
The problem is the form of my 'x'.  When I dig around in the
smooth.spline() code, it looks like it is relying on a call to IQR(x) which,
with data such as the object aa above, returns 0 because more than half of the
values are stacked on one value in the middle of the data.  The call to
smooth.spline() works if I add some random noise to the Xs, but I'd like to
be able to avoid that type of workaround.  Any other suggestions?

Thanks,
Nick

-- 
Nicholas G. Reich, Ph.D.
Research Assistant Professor
Division of Biostatistics and Epidemiology
School of Public Health and Health Sciences
University of Massachusetts, Amherst

413.545.4534
nick at schoolph.umass.edu
http://people.umass.edu/nick/
on twitter: @reichlab

On 15/02/2012 15:00, Nicholas Reich wrote:
> Hello.
>
> I'm getting an unexpected result when running smooth.spline().  Here
> is a simple example that replicates the error I'm getting:
>
>> aa<- c(1, 2, 3, 8, 8, 8, 8, 8, 8, 8, 8, 8, 12, 13, 14) bb<-
>> 1:length(aa) plot(aa, bb) smooth.spline(aa, bb)
> Error in smooth.spline(aa, bb) : need at least four unique 'x'
> values
>
>
>
> As you can see from the example, my 'x' clearly has more than 4
> values.  The problem is the form of my 'x'.  When I dig around in
the
> smooth.spline() code, it looks like it is relying on a call to IQR(x)
> which, with data such as the object aa above, returns 0 because more
> than half of the values are stacked on one value in the middle of the
> data.  The call to smooth.spline() works if I add some random noise
> to the Xs, but I'd like to be able to avoid that type of workaround.
> Any other suggestions?
Read the help page (as the posting guide asked of you) ...

      tol: A tolerance for same-ness of the ?x? values.  The values are
           binned into bin of size ?tol? and values which fall into the
           same bin are regarded as the same.

Works with tol = 0.1, for example.  IQR is only used in the default for 
that argument.

> Thanks, Nick


-- 
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