Re: fitting, step

Hi Maurice,


    Maurice> I am importing a histogram I made in another program. I use 
    Maurice> "define_counts" to make it into a spectrum I can fit with isis, and I 
    Maurice> assume a perfect response, fitting individual lines as Gaussians. I can 
    Maurice> define a Gaussian model with "fit_fun", preset parameter values to be 

I have done exactly that.

    Maurice> close to the best fit values using "set_par", and use "eval_counts" or 
    Maurice> "renorm_counts" to make sure that the function evaluates, and 
    Maurice> renormalize the line strength before fitting the centroid and width. So 
    Maurice> far so good.

    Maurice> However, when I do "fit_counts", it doesn't fit the data, or it barely 
    Maurice> changes the fit parameters. The one clue that something may be wrong is 

Is it very narrow?  If you use a large sigma, does that shrink?

How about if you use the subplex method?  (a single Gaussian should
fit w/ just about any method, but a subplex does a better search).

Have you excluded all other histograms (if any)?

Are the data integer valued? (that might matter if your uncertainties
are expected to be integers).

You can also use define_flux and fit_flux, if you have real-valued data.

    Maurice> that the "step" field is set to zero for all model parameters I want to 
    Maurice> fit, even if I explicitly set it using "set_par". I have checked this 

I'll let John and Mike reply to "step" - I've never used it.


Here's my quick test using an ideal case:

(w1,w2) = linear_grid( 1.0, 20.0, 1024);
y = exp( -(w1-10)^2 / 0.5^2 / 2) * 100;
dy = sqrt(y);
define_counts( w1, w2, y, dy ) ; 

set_par( 1, 1, 0, 0, 10000);
set_par( 2, 10.5, 0, 9, 12);
set_par( 3, 0.1, 0, 0.01, 1);
fit_counts(Ideal_ARFRMF);


-- Dave

David Huenemoerder  617-253-4283 (o); -253-8084 (f); http://space.mit.edu/home/dph
MIT Kavli Institute for Astrophysics and Space Research
70 Vassar St., NE80-6065,
Cambridge, MA  02139
[Admin. Asst.: Elaine Tirrell, 617-253-7480, egt_at_email.domain.hidden
----
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