[q] CERNLIB CHISIN equivalent?

Hello,

Where can I find CERNLIB CHISIN equivalent in Root?
I had to translate chisin function to calculate confidence
levels, see below. I found similar one in the TRolke class,
called “Chi2Percentile()”, but it seems to give un-expected
results, compared to PDG 2004 Table 32.3. I mean, it’s not
accurate (or I may be cinfused and the Chi2Percentile()
function is not supposed to be CHISIN equiv.)? Please look at my
implementation in the attachement. The output of my test
prog is the followings, as you can see CHISIN seems to give
exactly same answer as PDG.

regards,
Chul Su

PS - main script

go.C ----------

{
gROOT->Reset();

gROOT->ProcessLine(".L MyBayesUL.C+");

gROOT->ProcessLine(".x ul.C");
}

------ Output of test prog, “.x go.C” -----

PDG 2004, Statistics page 22 in pdg.lbl.gov/2004/reviews/statrpp.pdf
Table 32.3, nu_up

CERNLIB CHISIN test

n UP(90%) UP(95%)

0 2.30 3.00
1 3.89 4.74
2 5.32 6.30
3 6.68 7.75
4 7.99 9.15
5 9.27 10.51
6 10.53 11.84
7 11.77 13.15
8 12.99 14.43
9 14.21 15.71
10 15.41 16.96

TRolke::Chi2Percentile() test

n UP(90%) UP(95%)

0 2.30 3.00
1 3.89 4.85
2 5.47 6.44
3 6.87 7.75
4 8.23 9.39
5 9.55 10.78
6 10.84 12.15
7 12.12 13.50
8 13.38 14.82
9 14.63 16.13
10 15.87 17.42

MyBayesUL.C (13.8 KB)
MyBayesUL.h (1.08 KB)
ul.C (1.19 KB)

Hi,

I’ve checked the TRolke::Chi2Percentile function and it seems like there is indeed some problem of precision for small values of CL1 parameter. Something will be done to solve this issue.
Thanks for pointing out this problem,

Anna Kreshuk

Hi Anna,

I wondered if “chisin” equivalent function can be access from
out of a particular class, say, e.g. if it’s implememted in TMath
it will be better.

Thanks,
Chul Su

[quote=“anna”]Hi,

I’ve checked the TRolke::Chi2Percentile function and it seems like there is indeed some problem of precision for small values of CL1 parameter. Something will be done to solve this issue.
Thanks for pointing out this problem,

Anna Kreshuk[/quote]