Choosing a Flux Estimator

SCANPI provides several flux estimators: while it is clear that integrated flux estimators are better for extended sources, it is not the case that the template amplitude is always the best choice for point source fluxes. Some "rules of thumb" for choosing flux estimators are as follows:

  • For very strong point sources (> 20 Jy), the peak amplitude is a good estimator. In this case the template amplitude is affected by hysteresis, which will put "wings" on the source, which may elevate or tilt the template relative to the large-scale baseline.
  • For moderate point sources (1 to 20 Jy) the best estimators are the template amplitude, peak amplitude and fnu(t).(See material on SCANPI output for a description of the output tables and plots) Note that fnu(t) is obtained by integration within a fixed signal range. These three estimators should be in fairly good agreement for moderately strong point sources. Under some circumstances the template fit may be corrupted by background emission or noise, rendering it less valuable as a flux estimator.
  • For weak point sources (> 1 Jy at 12, 25, and 60 µm and > 2 Jy at 100 µm, SNR of a few) the best estimator is the template amplitude. Note that the cubic spline fitting can cause some "ringing" which tends to enhance weak point sources pushing their amplitude up by as much as 20% and their half-maximum width down. This is one of the main reasons for preferring template amplitude for weak sources.
  • For bright extended sources, fluxes are best estimated by fnu(Z), the integrated signal between the zero-crossings. However, if the SNR is less than about 10, it becomes difficult to identify reliable zero crossings, and fnu(t) may be preferable.

Diagnosing Unresolved Sources

To decide if a source is extended (in-scan) one of two tests can be used:

  • First, if the integrated flux is substantially larger than the peak flux (i.e., by several sigma)
  • Second, if the width of the signal is significantly larger than the expected width for a point source (see IPAC memo Statistical Characterization of SCANPI, Chapter I, Point Source Widths at 12 microns). This test is better carried out for the 25% width rather than the 50% width if the SNR is sufficiently high (> 20).

Choosing a Coadd to use

There is no general rule as to which of the coaddition methods produces the best flux estimates. The noise-weighted mean (1003) should minimize the noise, but is quite vulnerable to miscalibration. What is most important is that the results from 1001, 1002, and 1003 should agree to within about 1 sigma or so. If they disagree by more than this the individual scans should be examined more closely for peculiarities. Beyond that, the median is probably the most consistently "good" estimator of the three, chiefly because of the non-Gaussian nature of noise in the IRAS data.

<-- Previous Section   Next Section -->
(SCANPI Output) (FRESCO Overview)