We use PAW as one of our major, first-line, analysis tools. It works well at looking for correlations between the variables of an n-tuple, treating each entry as an independent "event". However it is very hard to get PAW to efficiently look at correlations between such events. SNO records events from very many independent sources, both internal and external, and correlations can arise between events from two types of mechanism:-
Examples are a radioactive hot spot or the Sun! Here the correlation is statistical, not on an event by event basis. The basic approach is to look for anomalies in phase space. The idea that an n-tuple file is a set of independent events i.e. is a mini-DST, is in keeping with this approach, so PAW is a good first step.
An example of this could be the 8Li source. In this case a single event gives rise to a further sequence of events, two or more of which are separately captured by the DAQ. Now there are strong correlations between sets of events within the mini-DST and here PAW is lacking. The fact that the correlated events come from a single point in space and time suggests that the first step is to localise the search in space and time. Any propagation quickly washes out the space correlation but time remains a very strong way to home in. Most interesting mechanisms will have times of a few seconds or much less, so that, even in a few hours of data, this becomes very powerful, given the fact that our data is time ordered.