|
Posted by Roger Hunter on October 11, 2002 at 20:01:16:
Hi all; This is not a test. It's an explanation of the Jones probability
calculation which I normally use. In each case, a "." marks a day without quakes, a"*" is a quake and the
"^^^^^" is a prediction window. There are 60 days in the time period.
Each set has predictions for "a quake in the 5 day window".
Set A has 28 quakes including 3 of mag. 7+ as the largest.
*
* *
* *
* * *
* * * * * **
*.*..*.*......*..*.**.*............*......*...........**....
^^^^^ ^^^^^ ^^^^^
Set A has 16 of 28 quakes in its' 3 windows including all 3 of the 7+ quakes.
Next we break the time period up into consecutive windows of the predicted size.
*
* *
* *
* * *
* * * * * * *
*.*..|*.*..|....*|..*.*|*.*..|.....|.....|*....|..*..:.....|....*|*....|
| | | | | | | | | | | |
Since there are 60 days, there are 12 such periods. Then we count the number
of window periods which have a quake in them. There are 9 with quakes.
Finally we calculate the probability as 9/12 = 0.750 or 75%
Set A has 3/3 hits for a 100% score so he's doing better than chance.
*
* *
* *
* * *
* * * * * **
*.*..*.*......*..*.**.*............*......*...........**....
^^^^^ ^^^^^ ^^^^^
Note that this ignores magnitude and location. That's because the database
only contains quakes of the predicted size and location.Discussion anyone? Roger
|
