[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Bug and Fix for Inverse Chi-Squared Method
From: |
Janes, Robert (Columbus) |
Subject: |
Bug and Fix for Inverse Chi-Squared Method |
Date: |
Thu, 9 Jan 2020 10:20:28 +0000 |
Hello,
I ran into a pretty serious issue with the cdf_chisq_Pinv method. Thank
goodness for open-source because I was able to come up with a simple fix.
In case Outlook considers a makefile sort of an executable I've copied it here
rather than attach it. You may have to add tabs back into it.
sample: testing
./testing
testing: obj/testing.o obj/imsl_c.o
gfortran -o testing obj/imsl_c.o obj/testing.o \
-L/home/rjanes/gsl/gsl-2.6/.libs -l:libgsl.a -lgslcblas
obj/%.o: %.f
gfortran -c -g -o obj/$*.o $*.f
obj/%.o: %.c
gcc -c -g -o obj/$*.o $*.c
The problem is that for large degrees of freedom the method will abort with an
error, "inverse failed to converge". In looking at the code and running test
cases I found that the limit of 32 iterations is too small. I found the method
failed many times once degrees of freedom got above 1000 or so. I increased
the iteration limit to 50 and did not find a single failure for a huge range of
degrees of freedom. I found successful iterations in this range were taking
28, 29 and 31 iterations. So hitting 32 maybe isn't much of a surprise. But
it suggests that maybe this isn't too efficient an iteration, at least for
large degrees of freedom. But I'll leave that to you experts.
Sorry about the Fortran code. I'm porting a large application from Intel
Fortran / IMSL to gfortran / gsl. Rather than change the base application
code, I just implemented the IMSL calls with gsl.
Of the attached files, testing.f is the test Fortran program. imsl.c is the
implementation of the IMSL routine with gsl. And gammainv.c is the updated gsl
code. The only change was 32 -> 50.
Let me know of you need anything else from me. Thank you.
Rob Janes
SGS Transportation
Senior Software Engineer
SGS – CMX
2860 N. National Road Suite A
US – 47201 – Columbus, Indiana
Phone: +1 - 812 - 378 - 7966
Fax: +1 812 - 378 - 3393
Email: address@hidden
Information in this email and any attachments is confidential and intended
solely for the use of the individual(s) to whom it is addressed or otherwise
directed. Please note that any views or opinions presented in this email are
solely those of the author and do not necessarily represent those of the
Company. Finally, the recipient should check this email and any attachments for
the presence of viruses. The Company accepts no liability for any damage caused
by any virus transmitted by this email. All SGS services are rendered in
accordance with the applicable SGS conditions of service available on request
and accessible at https://www.sgs.com/en/terms-and-conditions
gammainv.c
Description: gammainv.c
imsl_c.c
Description: imsl_c.c
testing.f
Description: testing.f
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- Bug and Fix for Inverse Chi-Squared Method,
Janes, Robert (Columbus) <=