IRAS Explanatory Supplement
V. Data Reduction
D. Point Source Confirmation
D.6 Overview of Weeks-Confirmation
Table of Contents | Index | Previous Section | Next Section
- Weeks-Confirmation Decision
- Weeks-Confirmation Position Refinement
- Weeks-Confirmation Statistical Processing
The last stage in the confirmation chain searched for observations of the same sources with time separations on the order of a week to six months. All hours-confirmed sources became input to the weeks-confirmation processor. Sources which were not confirmed were placed in the WSDB along with those which were, and no rejections were enforced until final catalog preparation. As each new region of sky was processed, the new sources were put unconfirmed into the WSDB, where they remained until that region was covered again, at which time the sources from the earlier coverage served as candidates to confirm the newer ones.
As each hours-confirmed source arrived for processing, a coarse window on the sky was used to select candidates from the WSDB. These may or may not have been weeks-confirmed already. The window was 10.3'.square, and the candidates were required to be separated in time by at least 36 hours from the source being processed. No flux tests were performed. Only a position test was used and if more than one candidate passed this test, only the candidate with the highest score was kept as a match. Because of the sequential testing, valid multiple sightings should have been confirmed pairwise as each new sighting entered the processing, removing the need to identify more than one correct candidate from the WSDB. When a choice had to be made, a counter was incremented for possible confusion. Typically only about 3% of the confirmed sources were diagnosed as potentially confused at weeks-confirmation.
If there were no acceptable candidates, the new source was placed in the WSDB. Otherwise position refinement and discrepancy statistical computations were performed.
D.6.a Weeks-Confirmation Decision
The confirmation decision algorithm was a position test applied to each pairing of the new hours-confirmed source with the candidates. The candidate position probability density function was mapped into the in-scan/cross-scan coordinate system of the new source. The test was based on the two-dimensional cross-covariance of the position probability density functions, evaluated at the observed nominal separation. The underlying principle was identical to the position test used at hours confirmation, and for relative twist angles of less than 4.6 degrees, the same algorithm was employed. Otherwise, a generalized implementation was used which was extensively complicated by the need to handle arbitrary relative twist angles between the axes of symmetry of the density functions. The details of this generalized version are discussed by Rolfe, Otake, and Fowler (1984). When the value of the cross-covariance fell below the threshold the candidate was released from consideration.D.6.b Weeks-Confirmation Position Refinement
Confirmed source sightings were processed for refinement of the position parameters by renormalizing the product of the probability density functions. This procedure was similar in concept to that of hours-confirmation, but was complicated by the generalization to arbitrary relative twist angles. When the relative twist angle was less than 4.6 degrees, the hours-confirmation algorithm was used. For larger angles, the refined density function was sufficiently Gaussian that it was satisfactory merely to assign a value of one second of arc to the uniform-error component.D.6.c Weeks-Confirmation Statistical Processing
As at hours-confirmation, histograms of the value of the threshold parameter for confined and non-confined sources were separately accumulated position data for both types were output for downstream analysis of the sky distribution of these events, and known source tracking was performed.
Statistical summations for computing the means and variances of position discrepancies were maintained over the SOP period and over periods of approximately 50 days. This was done for the observed sightings relative to each other, and separately for known sources relative to refined position parameters, where applicable. Good agreement between a priori error modeling and the observed dispersions was found.
Table of Contents | Index | Previous Section | Next Section