The nice new log=TRUE option in dpois appears to mess up the
case where lambda=0 (I was trying to calculate the likelihood
of a saturated model). Because the behavior is now always to
calculate the probability in terms of exp(log(prob)), there's
a test for lambda<=0 which really needs to be only lambda<0.
dpois(0:5,0)
ought to give
1 0 0 0 0
but gives NaNs instead.
Here's a patch to src/nmath/dpois.c that appears to work (although it
could ? be more
elegant).
*** dpois.c.orig Fri Feb 25 13:53:06 2000
--- dpois.c Fri Feb 25 14:16:19 2000
***************
*** 35,41 ****
warning("non-integer x = %f", x);
return R_D__0;
}
! if(lambda <= 0.0) ML_ERR_return_NAN;
if (x < 0)
return R_D__0;
--- 35,51 ----
warning("non-integer x = %f", x);
return R_D__0;
}
! if(lambda < 0.0) ML_ERR_return_NAN;
! if(lambda==0.0) {
! if (give_log) {
! if (x==0) return 0.0;
! else ML_ERR_return_NAN;
! }
! else {
! if (x==0) return 1.0;
! else return 0.0;
! }
! }
if (x < 0)
return R_D__0;
--please do not edit the information below--
Version:
platform = i686-unknown-linux
arch = i686
os = linux
system = i686, linux
status =
major = 0
minor = 99.0
year = 2000
month = February
day = 7
language = R
Search Path:
.GlobalEnv, Autoloads, package:base
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._