Converging on a good solution - checking residuals

At the end of any SPI data analysis, the spiros solution residuals provided in the log file must be checked. Do not trust any solution with residuals larger than 3 standard deviations. The only possible exception is when considering one of the few cases with relatively high signal-to-noise and signal-to-background ratio, such as the Crab or Vela. In these cases, large residuals may come from non-poissonian errors which are not taken into account by spiros.

The first analysis of any dataset will in most cases have too large residuals. The first thing to check is the table provided in the spiros log file containing the solution residual for each pointing. Although bad pointing are filtered in the first (pointing) steps, there might still be a few pointings with large residuals. The next step is to activate the AUTO filter of spiros by entering ``AUTO'' in the ``pointing-subset'' entry field. Re-run the pipeline and check the before mentioned table in the log file. It is important that no high residuals remain.

Inappropriate background modeling will also lead to large residuals. Try to improve the background modeling changing time scales and/or methods.

Check the list of detected sources. Large residual can be induced by the presence of bright source located just outside the FoV. Another case is when spiros is limited to a number of sources (through the No. of sources parameter) and that there is an additional source in the FoV. In this case, spiros cannot take this source into account and its counts are attributed to the other sources and/or to the background.

When all the above checks are performed, and if there are still large residuals, it may be time to let varying some sources. spiros in imaging and spectra modes assumes that the source fluxes are constant throughout the whole observation period. A good solution can therefore not be obtained if some of the sources are variable (see Sect. ).

Only if bad pointings remain despite all efforts you can remove them from the analysis. Note all bad pointings and enter the GOOD pointings in the pointings-subset entry field (see also Sect. ).