[R] many zeroes in rgamma ... what's going on?
Vincent Goulet
vincent.goulet at act.ulaval.ca
Sun Nov 18 23:24:09 CET 2007
Le dim. 18 nov. à 16:15, Gregory Gentlemen a écrit :
> Dear Dr. Dalgaard,
>
> Thank you for your insight! In fact, I did read the example
> documentation, however, it pretty much told me the same thing that
> my little simulation did: there is ALOT of point mass at zero.
>
> Is there any fix to this problem? Seeing that rgamma won't work
> accurately, if I wanted to plot a density of an inverse gamma
> distribution with small scale and shape parameters, how would I do so?
Package actuar has the {d,p,q,r}invgamma() functions (and quite a few
others), if this can be of any help to you.
> Peter Dalgaard <p.dalgaard at biostat.ku.dk> wrote: Gregory Gentlemen
> wrote:
>> Hello fellow R users,
>>
>> I wanted to view the density on the standard deviation scale of a
>> gamma(0.001, 0.001) prior for the precision. I did this as seen in
>> the code below and found that for some reason rgamma is giving many
>> values equal to zero, which is strange since a gamma distribution
>> is continuous. What is going on here?
>>
>> Thanks for any help in advance.
>> Greg
>>
> That sort of shape parameter gives a distribution with most of its
> mass
> squashed against the y axis, so random numbers underflow to zero. But
> why did you not read the Example section of help(rgamma)? The effect
> is
> clearly indicated there.
>
>>
>>> x1 <- rgamma(10000, shape=0.001, scale=0.001)
>>> sd1 <- 1/sqrt(x1)
>>> truehist(sd1, xlim=c(0, 1.5))
>>>
>> Error in truehist(sd1, xlim = c(0, 1.5)) :
>> 'nbins' must result in a positive integer
>>
>>> summary(sd1)
>>>
>> Min. 1st Qu. Median Mean 3rd Qu. Max.
>> 2.266e+01 9.311e+66 3.250e+153 Inf Inf Inf
>>
>> --
> O__ ---- Peter Dalgaard Øster Farimagsgade 5, Entr.B
> c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K
> (*) \(*) -- University of Copenhagen Denmark Ph: (+45)
> 35327918
> ~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45)
> 35327907
---
Vincent Goulet, Associate Professor
École d'actuariat
Université Laval, Québec
Vincent.Goulet at act.ulaval.ca http://vgoulet.act.ulaval.ca
More information about the R-help
mailing list