I have no idea and no real insight.
In general, regardless of method (PA, TA, FA, etcA), be careful of bias. This happens when you look for an event and try to figure out the pattern that preceded the event. The bias is if you don't look for all the cases of the pattern, then you may miss the times the pattern was not followed by the event. They do this in science -- first look for something, find something, then figure some way to test it out. If it's real, you should see a frequency or relationship that is consistent or better than random.
Quite often I've seen something, then figure out some way to code it so the computer searches for it, then the pattern-event relationship goes to 50% to 0% -- eg, it was not real, or it was random, and I saved myself a lot of trouble by discarding it.
In general, regardless of method (PA, TA, FA, etcA), be careful of bias. This happens when you look for an event and try to figure out the pattern that preceded the event. The bias is if you don't look for all the cases of the pattern, then you may miss the times the pattern was not followed by the event. They do this in science -- first look for something, find something, then figure some way to test it out. If it's real, you should see a frequency or relationship that is consistent or better than random.
Quite often I've seen something, then figure out some way to code it so the computer searches for it, then the pattern-event relationship goes to 50% to 0% -- eg, it was not real, or it was random, and I saved myself a lot of trouble by discarding it.
