Inverse Chi Square Distribution

[Luca Notini]

Dear group,

I am using the inverse chi square distribution vi, the problem I am having
with it is that when the degrees of freedom > 408 the output of the vi gives
the wrong result. When d.o.f > 408 the output is 10000 and it stays that way
as the d.o.f. is increased.

This is not the correct result as verified using equivalent functions in
other software packages.

Is there some kind of upper limit on the d.o.f. in LabVIEW which I have
exceeded?

Luca

[altenbach]

What is yout LabVIEW version. I have tried 7.0 and 7.1 and can confirm that the output is clipped to 10000. Your x-input must very close to 1 if you see clipping with d.o.f. of 408. What is your x input?

(even with x=0.999999, I need a d.o.f. of about 9000+ to see the clipping.)

This is not good. Have you reported it to NI as a bug?

[Luca Notini]

My input probability was 0.97 however the occurrence of an output value of
10000 seems to be independent on the input probability as you can see for
yourself on the attached vi.

In other word it does not seem to matter the input probability is, the
output is 10000 if d.o.f. if over 408!!!!

I am using LabVIEW 6.1.


Luca




"altenbach" wrote in message
news:50650000000500000034920100-1079395200000@exchange.ni.com...
> What is yout LabVIEW version. I have tried 7.0 and 7.1 and can confirm
> that the output is clipped to 10000. Your x-input must very close to 1
> if you see clipping with d.o.f. of 408. What is your x input?
>
> (even with x=0.999999, I need a d.o.f. of about 9000+ to see the
> clipping.)
>
> This is not good. Have you repo
rted it to NI as a bug?

[KileenC]

Hi,

I tested the Inv Chi Square Distribution.vi in LabVIEW 6.1, 7.0, and 7.1. It looks like there was a bug in LabVIEW 6.1 with degree of freedom values greater than 408, but the problem has been fixed in version 7.x. The VI seems to produce the correct output in LabVIEW 7.x and I have attached screenshots of the corresponding VI front panels as well as the actual VI in a zip file as well.

Let me know if you have any further questions. Thanks!

Kileen C.
Applications Engineer
National Instruments

[altenbach]

Kileen,

As I indicated, a similar bug still exists in newer LabVIEW version, it just got pushed to higher d.o.f. (~9500).
(See attached demonstration in LabVIEW 7.1.)
Just run it, slide d.o.f to higher values, and observe the clipping at x=10000.

This still does not look right!

[KileenC]

Hi altenbach,

Great catch! I've submitted this to the LabVIEW R&D team for further investigation.

Thanks again for all your help!

Kileen C.
Applications Engineer
National Instruments