Good question. I myself have been working on that one lately and wish I knew the answer. Let me have a few days to put a post together regarding that subject matter. My new computer programs are making it easier to generate the types of data needed to answer those questions.

I can say that it appears that in some cases, certain types of fairly strong warning signals may be generated as much as two years before the earthquake responsible for them occurs. And relatively weak signals may go back even further in time. Obviously that can complicate efforts to identify a time window for the quake.

I myself detected a strong signal about six months to the day before what I understand was the most powerful and destructive earthquake to occur in Japan since the 1995 Kobe quake. Signals which I believe might have been for the destructive June 22, 2002 quake in Iran were detected around April 21, and then again on May 21, 2002.