On 01/11/2012 08:52 PM, Robert T. Short wrote:
Before I file a bug report, maybe I should check that
I am not all wet.
If I recall my Bessel function theory,
besseli(n,x) = besseli(-n,x)
if n is an integer.
In octave 3.4.2 I get
octave:3> besseli(1,1),besseli(-1,1)
ans = 0.565159103992485
ans = 0.565159103992485
Looks good for n=1, but for n=10
octave:4> besseli(10,1),besseli(-10,1)
ans = 2.75294803983687e-10
ans = -1.40610343427994e-07
Not so good!
And for n=100
octave:5> besseli(100,1),besseli(-100,1)
ans = 8.47367400813812e-189
ans = 7.38028373423800e+170
BTW, the numbers for positive n agree to about 9 or 10 decimal
places with my tables. Also, BTW, the relationships for besselj
seem to work fine.
Anybody?
Bob
File the report. The n's
for n<-1 are wrong.
Michael
|