Hi everyone,
I have been using adaptIntegrate from the cubature package for a
multidimensional integral that has infinite variance (and so not
appropriate for Monte Carlo techniques). Most of the time it works but
sometimes (though not always) when I slightly increase the accuracy I want,
or increase the bounds of integration I get the following error:
REAL() can only be applied to a 'numeric', not a 'pairlist'
For instance if the calculation works with absError=1e-3 and takes
approximately 10 seconds then asking for absError=1e-4 I get the error
about REAL() within one second. So even though the computation should take
longer it doesn't seem to be starting. However later on when I ask for
absError=1e-4 it works.
The error is quite erratic so I don't understand what is going wrong or how
to prevent it happening.
There are three parameters that I sometimes change and each can cause this
error all of a sudden (but most of the time it doesn't so there doesn't
seem to be any consistency!)
- absError (a parameter from the adaptIntegrate function) I write this as
absError = 1e-x for some natural number x
- upperBound / lowerBound (a parameter from the adaptIntegrate function) I
write these as upperBound = c(a,b) for some real numbers a,b.
- a parameter that I pass to my function that is being integrated and is
just a real number
Once I defined a variable
A<-10
and then passed A to the function and got an error but when I tried the
exact same call but wrote 10 directly instead of using A it worked fine.
However later on using A again there was no problem.
Most of the time the error occurs when I slightly increase the desired
accuracy or the domain of integration.
Is this a known bug?
Thanks for any suggestions!
[[alternative HTML version deleted]]
______________________________________________
[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.
