If a source does not vary periodically, or if the data do not allow to search for a periodicity, it might be relevant to run a statistical test to check if the source is variable. Again, it is assumed that the source photons have already been extracted and are contained in the table sourx.tbl. To run a Kolmogorov-Smirnov test against constancy, a parameter file has to be created first via CREATE/PARFIL KS ks.par (see Table 7.6).
Table 7.6: Example of a parameter file for K-S test (UNBINNED)
If the parameter file is called ks.par, the K-S test is then run with the command
MAKE/KSTEST ksThe probability for variability is then displayed on the terminal, as well as some additional information. It is advisable to plot the result with the command
PLOT/KSTEST sourx_ksThis command will plot the observed and expected integral distribution of photons (normalized to 1.0), and the absolute value of their difference (normalized to 0.3). The threshold values for 90%, 95% and 99% confidence level for variability are also plotted.
In many cases, however, a direct test on the photon arrival times is not the best choice. For instance, a background subtraction or a series of corrections to the data (e.g. for vignetting or dead time effects) have to be applied before running the analysis. It is then possible to run a variability test in the case of binned data. If the binned light curve (in counts or count rate, corrected or not) is contained in the table binned_licu, and the relevant column to analyize vs. :TIME is called :MYRATE, a parameter file like the one in Table 7.7 can be used to run MAKE/KSTEST ks.
Table 7.7: Example of a parameter file for K-S test (BINNED)