[Duffy]

(03-05-2017, 05:39 AM)Skywise Wrote: For the record, my dates and times were determined using the random number function from the spreadsheet in Apache Open Office.

>> The association between our data sets was far greater than the association with Roger's.  Admittedly, first assumption seemed to indicate you had rode shotgun on my numbers, but this assumption is based only on this test.  There is insufficient data to determine a true conclusion, just as there is insufficient data to determine significance.  As Roger pointed out, something positive has to occur with any hypothesis before it can be credited.  Speaking of Roger, I feel I inadvertently put him on the spot yesterday, it was genuine scrutiny of the data, "not" accusation ... out of the air and Apache Open Office is good enough for me !   

Duffy, the point of random trials in any statistical analysis is to see if randomly chosen data gives statistically similar results to the test data. You have a set of a data that seems to correlate to earthquakes. So Roger and I chose random data for you to test. The random data gave similar results. The implication is that your method does no better than random data, therefore is not statistically significant.

>> As mentioned, the sample is to small to determine a definitive answer, but if I was to give judgement on the available data, I would have to agree statistics do not show significance. 

But as Roger astutely pointed out, the total data is quite small and this leads to a larger margin of error.

>> Records from last year shows my signal source averages between 4 - 6 data lines per month, except December/ January and June/July when that number reached 12 -14. The period indicates correlation with the sun's most Northerly and Southerly position. I miss-interpreted their meaning at first, and used them as longitude indicators for prediction, thus I ended up at Lawrence Island, Alaska and the Siberian Sea etc.  I even enquired here about historic activity for a location in the South Pacific, far from any tectonic borders.  I recently used this source to predict events on several correct locations ... that's why I took the step of testing it here. I cannot prove the source as credible in said predictions because it would be "after the fact". I made this mistake with Hawaii because it is easy enough to get a few dates and times after the event has occurred.  My next best shot is June/ July if you and Roger are game for a rematch ? 

There's a saying, "There's lies, damned lies, and statistics." Statistical analysis can be a very tricky thing and many a scientist has been tripped up by subtle mistakes. Roger understands this and this is why he is so critical in his efforts to understand exactly what you are doing and to make sure his software is working right.

>> My father-in-law was director of electronics at a large manufacturing company, he use to write his own software to assist with production. Sometimes we would check with him before visiting, because problems with his software did not make for a pleasant stay ... I know nothing of programs, but I do know how Roger must sometimes feel.


Duffy