RE: Chi2 calculation for rebinned spectra in ISIS

Hi Joey, Mike,

so indeed this is the solution to the problem. If I use 'set_rebin_error_method(1,"quadsum");', as suggested by Joey, I get exactly the same chi2 and parameters in ISIS and XSPEC.
And yes, the 'problem' with XSPEC is that it is starting with a lot of bins with count 0 and error 1.866 (as Mike points out), so when it adds those errors in quadrature it gets to very high values.

Many thanks to you both and all who replied before!

Maria

________________________________
From: owner-isis-users_at_email.domain.hiddenSent: 30 April 2014 18:08
To: Joey Neilsen
Cc: isis-users_at_email.domain.hiddenSubject: Re: isis-users: Chi2 calculation for rebinned spectra in ISIS


On Apr 30, 2014, at 11:48 AM, Joey Neilsen <neilsenj_at_email.domain.hidden
It looks like dmgroup adds the errors of the grouped bins in quadrature. I believe ISIS defaults to Poisson errors upon rebinning. If you do:

set_rebin_error_method(1,"quadsum");
rebin_data(1,25);

and set the fit parameters to your best fit in XSPEC, do you get the same chi^2 in ISIS?

And for the unbinned HETG files, we default to the Gehrels chi^2 statistic == 1  + sqrt(counts+0.75) ~ 1.866 for all the 0 counts channels.  So, that definitely increases the size of the error bars for the XSPEC case.  I think that's the solution, at least for ISIS.  You will definitely want to shrink those error bars for XSPEC.

-Mike



----
You received this message because you are
subscribed to the isis-users list.
To unsubscribe, send a message to
isis-users-request_at_email.domain.hiddenwith the first line of the message as:
unsubscribe