Why does the GALFIT iteration, i.e. “countdown\, clock keep bouncing up and down?
• GALFIT has to know when to stop iterating. The ideal situation is when Chi2 changes by precisely 0, but in practice the reality is much more complex and this situation is just one of several. Many times, a non-linear least-squares algorithm would converge on a solution quickly but put on a slow break as it asymptotically approaches the minimum. At this point, when the Chi2 continues to decrease, but no longer appreciably to be statistically significant, it is time to call it quits. But there are also several other reasons why the solution may slow, the most obvious of which is that it has reached a long, very slowly descending, plateau, beyond which the Chi2 valley drops away quickly again. That can happen if, for example, the sigma image is very bad, or the parameter space may be very complicated such as when there are neighboring contaminations, or because something is inherently a hard fit. Therefore, quitting at the first sign of a slow Chi2 decline is a bad idea. On the other ha
