GDAAS ALGORITHMS
                           ================
          V.Belokurov, F.Figueras, D. Evans, L.Eyer, P.Popowski

Here, we list the different decisions, actions, which have been taken.

1) In the Cambridge meeting:
----------------------------

For the moment, due to the very high G photometric precision, it has been
decided to use the G band to detect the variability on a per transit basis
for all sources, that is to take the average of one passage over the 11
CCDs of the Astro field.

Some studies of very short term variability should be also explored (that
is on a per CCD analysis - some of the algorithms developed for the per
transit analysis could be reused).

On this G-band data we perform a number of statistical tests. The tests
proposed by Laurent were chosen, but other tests can be added without
problems.

For these tests two aspects are important:
  - They should be formulated in a cumulative way, so that during the
    mission, "tables" of the status of the variability for all sources are
    available. The tables are dynamic and evolve during the mission.

  - P-values should be computed. This is through the method of hypothesis
    testing and the notion of P-values so that stars with different numbers
    of measurements and different noise levels can be compared.

It was noted that the tests as they are currently formulated have been
checked with respect to Hipparcos data (note the P-value of kurtosis seems
to have not been well calibrated).


2) In the Barcelona meeting (between Laurent and the Barcelona Group):
---------------------------

It is also important to have the tests developed for the end (mid)
mission, so that the algorithms are also coded in "non-cumulative" way.
This is somewhat objected by Dafydd. However other tests which are not
easily computed in a cumulative way can be introduced here.


3) September GDAAS Algorithm:
-----------------------------

It is the Cambridge proposal which is currently implemented.


4) In the Paris meeting:
------------------------
    
There has been no opposition to the addition of a period search algorithm
(Lomb-Scargle algorithm). This step is important, so that we have an idea
of the computation load.
    
    
TASKS:
-----
Laurent: To pass the algorithm to Vasily (done) and to Piotr
Dafydd : To screen all algorithms (underway)
Laurent: To write an introductory document on the algorithms (underway)
Laurent: To introduce period search algorithm
Laurent: Test should be conducted to extract the data from GDAAS2
Laurent: To study SDSS photometry for some tests about multicolour
         variability


Laurent & Dafydd,
The 1st November of 2004