I am having trouble with the functions for the gamma distribution for
small parameter values, say shape=.001 and rate=.001, which has mean 1
and variance 1000. For these values I get
> qgamma(c(.01,.05,.25,.50,.75,.95,,.99), .001,.001)
[1] 1.000000e-051 1.000000e-051 1.000000e-051 5.244206e-299
6.470515e-123 2.973588e-020
[7] 2.425943e-002
which are clearly wrong. For values exceeding .01, the functions appear
to work fine. Is it documented someplace what the lower bounds of the
parameters must be to maintain accuracy?
Joe Lucke
jlu100+.vcf
Description: Card for Joseph F Lucke
|