I think you will find the following informative.

Check .ps files in /data/L3/adam/detecttest/analysis.

DETECTION EFFICIENCY

"eff.ps" shows detection efficiency, i.e. how often an isolated source
that was put in the simulation was in fact detected.

There are two pages showing the same results: on the first page
individual plots show the results for same strength sources, on the
second page the individual plots show same simulation exposure times.

Sources with N_evt of 30 and higher are always detected. Note that for
those sources we only have 3ks and 100ks simulations (i.e. extreme
cases). But given 100% success rate for 100ks (worst case), it makes
no sense to do more of them.

RESOLVING CLOSE SOURCES

"res2.ps" and "res4.ps" show percentage of cases when two sources
separated by 2 and 4 arcsec, respectively, were resolved.  Important
note: this is independent of whether anything was actually detected or
not (but in almost all cases something was detected; note that for
large off-axis angles this effectively becomes detection of "single"
sources with twice the number of photons).

We do not have many simulations for those cases, and some of the
curves are rather wiggly.

FALSE SOURCE RATE

"fsr.ps" shows the number of false sources per simulation found by
wavdetect in one arcmin annuli (keep in mind that the area covered by
the annuli grows linearly). The x-range is shorter here than in other
plots because for large off-axis angles the simulations show lots of
false sources caused by detector edge effects, and I needed to exclude
those.

The top left panel includes absolutely everything, ie. even obviously
bogus sources. In a quick attempt to weed out those bogus sources, I
excluded sources with significance (as given by wavdetect) smaller
than 3 (top right panel) and with net events smaller than 3*error
(bottom right panel).

The turnover around 6 arcmin off-axis is a result of the fact that
this is where PSF really baloons and wavdetect (wrecon, to be precise)
gets better in rejecting "sources" that really are local background
fluctuations based on their size alone.

AD, 12/12/2002