[R] Beyond double-precision?

joaks1 joaks1 at gmail.com
Sat May 9 09:13:13 CEST 2009


Yes, all of the numbers are positive.  I actually have a Bayesian posterior
sample of log likelihoods [i.e. thousands of ln(likelihood) scores].  I want
to calculate the harmonic mean of these likelihoods, which means I need to
convert them back into likelihoods [i.e. e^ln(likelihood)], calculate the
harmonic mean, and then take the log of the mean.  I have done this before
in Mathematica, but I have a simulation pipeline written almost entirely in
R, so it would be nice if I could do these calculations in R.

If R cannot handle such small values, then perhaps there's a way to
calculate the harmonic mean from the log likelihood scores without
converting back to likelihoods?  I am a biologist, not a mathematician, so
any recommendations are welcome!  Thanks! -Jamie


spencerg wrote:
> 
>       Are all your numbers positive?  If yes, have you considered using 
> logarithms? 
> 
>       I would guess it is quite rare for people to compute likelihoods.  
> Instead I think most people use log(likelihoods).  Most of the 
> probability functions in R have an option of returning the logarithms. 
> 
>       Hope this helps. 
>       Spencer
> 
> joaks1 wrote:
>> I need to perform some calculations with some extremely small numbers
>> (i.e.
>> likelihood values on the order of 1.0E-16,000).  Even when using the
>> double() function, R is rounding these values to zero.  Is there any way
>> to
>> get R to deal with such small numbers?
>>
>> For example, I would like to be able to calculate e^-10000 (i.e.
>> exp(-10000)) without the result being rounded to zero.
>>
>> I know I can do it in Mathematica, but I would prefer to use R if I can. 
>> Any help would be appreciated!
>>
>> Many Thanks in Advance!
>>
> 
> ______________________________________________
> R-help at r-project.org 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.
> 
> 

-- 
View this message in context: http://www.nabble.com/Beyond-double-precision--tp23452471p23457955.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list