An example

For the example I'll use the 8C1435+635 measurement at 850$\mu$m that was discussed in §4.2. Here the whole data set has been reduced with sky removal, concatenated and despiked using sigclip (see Figure ). Note that sigclip must be used rather than drawsig because it produces an output file devoid of spikes. I renamed this file 8c_clip for brevity.

Figure: A 48 minute (160 x 9/2.0 jiggle patterns) integration on 8C1435+635. The sigclip command has been used to produce a dataset despiked at the 3$\sigma$ level.

The kstest command is part of the Starlink KAPPA package and takes a variety of parameters. In this document I'll just discuss those relevant to the basic KS test as applicable to photometry data. A full description of the routine can be found in the KAPPA manual (SUN/95).

I'll split the dataset into subsamples of 20 integrations (the second subsample will contain only 19 points because sigclip removed one spike). In this example, the input and output file names are specified on the command line but this need not be the case. Note that it is, however, necessary to give the number of samples and the probability in the same way, otherwise the default values of 3 and 0.05 are assumed respectively. A maximum number of 20 subsamples is allowed by kstest.

% kstest in=8c_clip out=8c_ks nsample=8 limit=0.05
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 19 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
 
                  Probability        Max. Sep.
--------------------------------------------------------------
 cf. with   2 :   0.288655E+00       0.300000E+00  (Accepted)
 cf. with   3 :   0.244933E+00       0.270513E+00  (Accepted)
 cf. with   4 :   0.981682E+00       0.116102E+00  (Accepted)
 cf. with   5 :   0.924850E+00       0.132279E+00  (Accepted)
 cf. with   6 :   0.288812E-01       0.344444E+00  (Rejected)
 cf. with   7 :   0.380108E+00       0.215151E+00  (Accepted)
 cf. with   8 :   0.583820E+00       0.181092E+00  (Accepted)
 
Number of subsamples rejected: 1
Coadded result is 4.8022197E-5 +/- 8.0190266E-6

Subsample 6 which contains many more points with a signal greater than zero compared to the other subsamples is rejected (see Figure ). If you wish to view the truncated dataset then 8c_ks.sdf can be displayed with qdraw in the usual way.

If the first sample is anomalous then we can repeat the test in reverse order by applying the KAPPA command flip to the original dataset, 8c_clip;

% flip 8c_clip 8c_flip

More generally, any subsample can be specified with a NDF section [8]. In the above example, subsample 7 is 8c_clip(121:140). The kstest command can then read in an ascii file containing a list of NDF sections, such as,

% cat kstest.dat
8c_clip(121:140)
8c_clip(1:20)
8c_clip(61:80)
8c_clip(81:100)
8c_clip(21:40)
8c_clip(101:120)
8c_clip(141:160)
8c_clip(41:60)
% 

The following command then performs the KS test using this file;

% kstest '^kstest.dat' out=8c_ks limit=0.05
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 19 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
Data for "8C1435+635" contain 20 points, of which 20 are good.
 
                  Probability        Max. Sep.
--------------------------------------------------------------
 cf. with   2 :   0.497342E+00       0.250000E+00  (Accepted)
 cf. with   3 :   0.767976E+00       0.175000E+00  (Accepted)
 cf. with   4 :   0.860007E+00       0.150000E+00  (Accepted)
 cf. with   5 :   0.444130E+00       0.212500E+00  (Accepted)
 cf. with   6 :   0.100815E-01       0.384849E+00  (Rejected)
 cf. with   7 :   0.835178E+00       0.146970E+00  (Accepted)
 cf. with   8 :   0.189298E+00       0.253361E+00  (Accepted)
 
Number of subsamples rejected: 1
Coadded result is 4.8022197E-5 +/- 8.0190266E-6
 
% 

Note that the NSAMPLE parameter is now redundant and also that single quotes must surround the input file name. The final result in this case is unchanged, of course.