POLE(the software is from https://github.com/ftegenfe/polepp ) problem

ROOT
1

Hi !
I am having some problems of POLE software. This POLE is from the github and the developer is from CERN, so I want to ask some helps in root forum.

This POLE used to calculate upper limit by using Feldman-cousins methods including systematic error, thus I download the POLE and try to run some example:

the run command is:

./polelim --effmeas 1 --effdist 2 --effsigma 0.42 --bkgmeas 0.59 --bkgdist 2 --bkgsigma 0.38 --nobs 0

however the output is below, it seems wrong but I can’t know how to solve this problem:

 *Test hyp. step    : 0.01
----------------------------------------------
 Step mu_best       : 0.01
 Max N, mu_best     : 20
----------------------------------------------
 Method             : FHC2
----------------------------------------------
 Verbosity          : 0
----------------------------------------------
 Parameters prefixed with a * above are not   
 relevant for limit calculations.
==============================================

------------------------------------------------------
- ERROR: limit calculation failed. Possible causes:
- 1. binary search threshold too large
- 2. precision threshold too large
- 3. precision in integrations (eff,bkg) not sufficient
- 4. poisson table insufficient precision -> try without tabulated poisson
- 5. minimum probability in belt calculation too large; min = 0.001
-
- Results:
-    probability (should be ~ CL)  = 0.890523
-    lower lim norm (should be ~1) = 1
-    upper lim norm (ditto)        = 0.89976
-    lower lim                     = 0
-    upper lim                     = 2.29625
------------------------------------------------------

So is there someone is familiar with this POLE software or someone who knows the developer of POLE: Fredrik Tegenfeldt ? I want to contact with the developer and fix the problem as soon as possible. Thanks !

2

You could try an email: “Fredrik.Tegenfeldt_at_cern.ch”

3

Thank you！ I have sent an email to Fredrik last week but I don’t receive a reply

4

Welcome to the ROOT forum.

Try to post an issue on the github repository.
BTW which github repository are you talking about?

5

There is no Fredrik Tegenfeldt in the CERN Directory…

6

The repositry website is in the title of this topic.

7

Ok, it a little bit wired :disguised_face: :sweat_smile:

8

I see the repo now, Sorry I missed it in the title.
As @bellenot said this guy is not visible in the CERN directory.
Maybe he was 17 years ago when the software was created.
I think the best is to post an issue on the GitHub repository.

9

Ok thank you
I can try

10

BTW, is there something like this in ROOT to calculate the upper limit including systematic error using Feldman-cousins methods ?

11

https://root-forum.cern.ch/search?q=Feldman-cousins
https://root.cern/doc/master/FeldmanCousins_8C.html

etc…

12

Hi,

I was somehow tagged in this post.

Try using the simple case:

./polelim --effdist 0 --bkgdist 2 --bkgmeas 1 --bkgsigma 0.38 --nobs 0

If one reduces the bkgmeas to 0.9 then that error comes and also similar by increasing the bkgsigma…
It seems to be a bug, try varying the values slightly then for such small values the UL won’t change drastically. You can follow the the paper: Phys.Rev.D57:3873-3889,1998 for comparing the UL values…

13

Yes. POLE can work well just in some values. However it seems nobody maintain POLE program now thus I decided use RooStat to calculate upperLimit. Thank you

14

Thank you for your advise. I am trying to use RooStat. Thank you very much.

15

Hi Debashis!
I meet some problems with Standardhypotestinvdemon.C. Below is a post:
https://root-forum.cern.ch/t/problem-of-using-standardhypotestinvdemo-c-to-calculate-the-upper-limit-of-feldman-cousins-methods/54745

Could you help me ? Thank you very much

16

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.