01-26-2015, 10:49 PM
(01-26-2015, 08:48 PM)Roger Hunter Wrote: Ideally I'd want them before the quakes happened so you couldn't be accused of selecting only good ones for testing.
Agreed, but I think we can still evaluate past data if made available. Although cherry picking is always a concern, I vote for innocent until proven guilty. That is, let's not assume the data is tampered with unless there's good reason to think it was. Just because some other quake predictors have played with the data doesn't mean Duffy has.
(01-26-2015, 08:48 PM)Roger Hunter Wrote: There are some problems with this, in that you aren't specifying date, mag. or location which makes it difficult to say that a given signal matches a particular quake.
I'd like to expand on this to make sure Duffy understands where you're coming from. To evaluate a possible prediction or precursory phenomena, it makes the math much easier and the analysis more robust if it can be said that a given precursor results in a quake of a given location, date/time, and place. All three of these variables can be quantized, and all three are part of the evaluation process.
The difference is in saying, "there's going to be a mag 8 quake in Japan" vs "there's going to be a mag 8 quake on the Ring of Fire". If the quake happens in Chile, the first statement is clearly false, but the second is true. However, the area of the Ring of Fire is much much larger than just Japan, therefore the second statement, although true, carries much less significance than if the quake had occurred in Japan.
Ideally we want to eliminate any fuzziness to the variables. If the dart lands in the outer ring of the dartboard, it's NOT a bulls eye. But if someone says "near the bulls eye", it then becomes an argument of "how near is NEAR"?
However, even saying that a quake of mag 7+ will occur somewhere on the planet within 5 days of the onset of a SID can still be evaluated. It just carries less significance per event, and therefore we'd need to study a much longer time period to tease out the true significance.
Anyway, an evaluation can still be done in most circumstances, and I agree with Chris that this VLF data sounds very interesting. It's hard data, replicable, testable, and definable. It's not fuzzy.
Brian