Photometric observations of faint sources often require that the source be observed for long periods, during which the atmospheric conditions and/or telescope related parameters such as pointing and focus may vary. These effects are particularly noticeable around sunrise and sunset, for example. The consistency of the entire data set can be investigated with the use of a two sample Kolmogorov-Smirnov (KS) test. Again, the method adopted for the SCUBA data reduction package is analogous to that used previously by COADD for UKT14 [7].
The data set is split into subsamples that are specified by the user. The size of these subsamples clearly depends on the total number of photometric points and some experimentation is required but, for example, if the data set consists of 100 points then subsamples of 20 would be a sensible choice. The first subsample is compared with the second and is `rejected' if the probability that the two are statistically different is lower than a user specified tolerance. If the KS test returns 0.1 then this indicates that there is a 90% probability that the two samples are different. If the first two subsamples are consistent then they are concatenated and compared with the third etc, otherwise the first subsample is compared with the third and so on. A problem arises if, for some reason, the first subsample is different from each of the others since in this case most of the data set will be rejected. If this happens it is necessary to repeat the process but this time select a different subsample against which the others are to be compared. In general, it is good practice to change the order and repeat the test.
The best way to illustrate the whole process is with an example using real data.
The SCUBA photometry cookbook