Re: puzzlin Chandra proposal thread

Hi Saku - 

I think you are correct.  And assuming the appropriateness of the
statistic and model, ISIS is doing the right thing.  This more of a
statistics question.  As I understand this thread, it intentionally
constructed a simulation with no feature, then asks the question, "how
big a feature could be there at this noise level without me knowing
it?"  So I believe I've misinterpreted the conf result before (and
need to check an elementary statistics text!).

For a priori known flux of 0.0, we fit 1.5e-7, but it could be
2.6e-5 or 0, to 90% confidence (there is probably a problem w/ the
lower limit, since flux is constrained to be >0).

So I don't see how the thread as written answers the original
question posed.   If your line flux is greater than 2.6e-5, then you
can say there is a 10% or smaller probability that it is due to noise;
but I don't think you know the detection significance.

--Dave



In isis-users, you wrote:
>
>The confidence limits should be on the value determined by the fit.
>
>In this thread after running fit_counts if you list the parameters:
>
>isis> list_par;
>phabs(1)*(powerlaw(1)+gauss(1))
> idx  param             tie-to  freeze         value         min         max
>  1  phabs(1).nH            0     1             0.32           0      100000  
>10^22
>  2  powerlaw(1).norm       0     1            0.047           0       1e+10  
>  3  powerlaw(1).PhoIndex   0     1             1.85          -2           9  
>  4  gauss(1).area          0     0     1.503894e-07           0           1  
>photons/s/cm^2
>  5  gauss(1).center        0     1            12.13           0           0  A
>  6  gauss(1).sigma         0     1             0.08           0           1  A
>
>
>so the value of the parameter is 1.5e-7 and the confidence limits are 
>0 and 2.6e-5.  Doesn't that mean that the value is consistent with 0?
>----

----
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