next up previous contents
Next: Oversampling Up: PSS Reference Previous: Fitting Statistics

Background Rescaling

PSS works assumes the background model to be correct. For some images however, the model may not be of good quality -- it may have been obtained using a simple function fitting approximation for example. If the Cash statistic is being used, PSS will offer the following option in expert mode.

   RESCALE - Re-scale background estimate /NO/ > Y
Selecting this option invokes a derivative of the Cash statistic described in Section B.1. Both the flux and background scaling factor are optimised at every point using this statistic (ie. at every point within the slice in SEARCH mode, or every test position in PARAM mode).

An initial guess for the scale factor is requested.

   ISCALE - Initial guess for background scale parameter /1/ >
Background rescaling is useful where the background is known to be roughly right but needs fine tuning. It will not correct background models containing unrepresentative gradients on spatial scales comparable with the PSS box size.

Needless to say, given the complexity of the algorithm, background rescaling is significantly slower than the standard Cash statistic. The exact time penalty depends how well the background model supplied estimates the true background .



Asterix
Tue Oct 7 12:03:17 BST 1997