1. Simulated count rate profiles for 100,000 c/s/SC from a Gaussian source superposed on model profiles using the same phase_map_ctr and Gaussian source.

2. Contours of the Cash statistic for model Gaussian profile and Simulated profile. 1000 cts/s/SC, widths=[0.1,0.1], SIM_XY=[31,31] Note that the Chi-square statistic (dotted) has multiple extrema, displaced from the Cash statistic minimum. This is typical behavior for count rates of this order. The 0.5,0.5 arcsec displacement of the Cash statistic minimum from (0,0) may be due to pixelization in the source simulation (hsi_image), although one of the signs appears to be wrong.

3. Total Cash statistic for model Gaussian of guessed widths 1-100 arcsec, simulated width=8x8 arcsec. The Cash statistics for the fine subcollimators are underweighted in this test, but note that the minimum lies near the true width of 8. The Cash statistic should be weighted in proportion to the number of bins for a proper measure of goodness of fit.

4. The Cash statistics computed separately for each detector as a function of guessed size (1-100 arcsec), for the 8x8 arcsec simulation of the previous figure. There are 3 things to note: The finest subcollimators over resolve the source and give little size information; The coarsest subcollimators put only an upper limit on source size Only the subcollimator with resolution ~3 x source width shows a significant minimum

5. Cash statistic for simulated 4x8 arcsec Gaussian as function of guessed tilt angle. As before, there are 3 things to note: The finest subcollimators over resolve the source; The coarsest subcollimators under resolve, and give no tilt information; Only the subcollimators with angular resolutions from ~ 2 to ~5 times the widths give a minimum near the correct value of tilt (45 degrees). This again indicates that the parameter search must "know" about under- and over - resolution.

Last modified May 22, 2000.