# Re: ISIS and its fit statistics

From: John Houck <houck_at_email.domain.hidden>
Date: Sat, 7 Feb 2004 16:54:51 -0500
On Fri, Jan 30, 2004 at 10:12 -0800, FIll Humphrey wrote:
> 2. For both the Cash and \chi^2 fit statistics, to my very inexperienced
> reading of the code, neither seem to take into account the errors on the
> background you read in. Is this correct? I've noticed tangible differences
> between fits done with XSPEC and ISIS, and was wondering if this is the
> root of the discrepancy. If the background's not taken into account, would
> it be possible to deal with it ? la XSPEC?

My original answer to this question was incorrect -- the
background uncertainty _is_ included.

You can verify this empirically by computing the fit-statistic,
then unassinging the background and noting the change in the
statistic value.

To see it in the code, look at
histogram.c:Hist_copy_noticed_data
which is called from

Could you quantify what you mean by "tangible differences"
between fits done with xspec and isis?

Are the differences within the range of expected statistical
errors?

Because the two codes do not use identical algorithms and
identical convergence criteria its not surprising that the
answers are not identical.  But if you see differences which
are statistically significant, I'd like to understand that.

Thanks,
-John
----
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

Received on Sat Feb 07 2004 - 16:55:16 EST

This archive was generated by hypermail 2.2.0 : Thu Mar 15 2007 - 08:45:50 EDT