[Roger Hunter]

(03-15-2015, 01:01 AM)Skywise Wrote: Test the null hypothesis. Try randomizing date/times of the quakes. Keep the random date/time within the same time span.

Do you have a reason to think the average number is not a good null hypothesis?

Is attaching the image the right way to insert it? Seems to work ok.

Roger

[Skywise]

(03-15-2015, 01:25 AM)Roger Hunter Wrote:

(03-15-2015, 01:01 AM)Skywise Wrote: Test the null hypothesis. Try randomizing date/times of the quakes. Keep the random date/time within the same time span.

Do you have a reason to think the average number is not a good null hypothesis?

The idea is, if there is no relationship between sun-moon angles and earthquake timing, then randomizing the quake times should result in a qualitative if not quantitatively similar graph.

Flip a penny 100 times. I would not be surprised by seeing the 10 heads in a row. Even 20. Sometimes you get lucky.

(03-15-2015, 01:25 AM)Roger Hunter Wrote: Is attaching the image the right way to insert it? Seems to work ok.

Looked good to me. To be honest, the graph didn't look terribly "peaky" to me so doesn't seem to me that the peaks are particularly significant.

But as we know, numbers don't lie.

Brian

[Roger Hunter]

(03-15-2015, 02:25 AM)Skywise Wrote: The idea is, if there is no relationship between sun-moon angles and earthquake timing, then randomizing the quake times should result in a qualitative if not quantitatively similar graph.

Flip a penny 100 times. I would not be surprised by seeing the 10 heads in a row. Even 20. Sometimes you get lucky.

Ok, here's 6000 random dates from 1900-2000.

Equally significant. Shouldn't be.

Roger

[Skywise]

(03-15-2015, 03:42 AM)Roger Hunter Wrote:

(03-15-2015, 02:25 AM)Skywise Wrote: The idea is, if there is no relationship between sun-moon angles and earthquake timing, then randomizing the quake times should result in a qualitative if not quantitatively similar graph.

Flip a penny 100 times. I would not be surprised by seeing the 10 heads in a row. Even 20. Sometimes you get lucky.

Ok, here's 6000 random dates from 1900-2000.

Equally significant. Shouldn't be.

Roger

Look at it the other way. The real data is equally INsignificant as 6000 random dates.

I would then say that the proposition under test fails.

Brian

[Roger Hunter]

(03-15-2015, 04:21 AM)Skywise Wrote: Look at it the other way. The real data is equally INsignificant as 6000 random dates.

I would then say that the proposition under test fails.

Brian

What I think this shows is that the average is not a good null hypothesis. These are mag 6+ quakes. They don't happen all that frequently so the sun-moon angle is erratic.

6000 random dates in 100 years is also erratic and gives a similar answer; the distribution does not fit the expected line.

Roger

[Roger Hunter]

(03-15-2015, 12:32 PM)Roger Hunter Wrote: What I think this shows is that the average is not a good null hypothesis. These are mag 6+ quakes. They don't happen all that frequently so the sun-moon angle is erratic.

6000 random dates in 100 years is also erratic and gives a similar answer; the distribution does not fit the expected line.

Roger

Further tests with larger samples shows that the distribution converges to the average with enough cases. VERY large numbers.

Smaller sets just don't work. What's important is the pattern; are certain angles found more often with quakes and the answer is no.
Different samples give different answers.

Roger