+- Earthwaves Earth Sciences Forum (http://www.earthwaves.org/forum)
+-- Forum: Earthwaves (http://www.earthwaves.org/forum/forumdisplay.php?fid=16)
+--- Forum: Miscellaneous (http://www.earthwaves.org/forum/forumdisplay.php?fid=15)
+--- Thread: A Fair Test 8 / 2 / 2017 (/showthread.php?tid=697)
A Fair Test 8 / 2 / 2017 - Duffy - 02-08-2017
I thought I would review the test again, that I recently carried out between 16th January and 28th January, the day of new moon. It was a simple set up, you choose a period of time to run the test, you select 24 locations, and then see how many times the locations are linked to times of events.
My method was to use a large flat screen TV connected to my laptop. I represented my locations with crosses on the screen, and I used a 2D online map program from http://www.timeanddate.com to best intemperate global terminator positions. When one of the terminator thresholds was in contact with a cross, at the time a 5+ event occurred, I googled the location of the cross down to a specific feature on the google map. I then found this same feature on http://www.suncalc.net, which utilises the same google system. It allows you to select features or locations, and one click provides you with the necessary terminator information. At the end of each day (my time), I would post any contact results on the forum so that any individuals wishing to monitor this, would either come to the same conclusion, or determine any discrepancies.
The only agenda with this test was to determine if the signal providing the location data, had any significance with respect to earthquakes. I did not do this to prove association, nor am I claiming this to be any kind of trigger (though I am exploring this possibility). The fact is, I have come across a strange phenomena that does not seem to have a definite answer, but there are a couple of points to consider about this test.
I ran the test in real-time ... which means just like everybody else watching its progress, I would have no idea of the eventual outcome. My only advantage was that I had done this before ... so my definition of "chance" with this test is, I took a big one posting it here. The mere fact that Roger had trouble understanding this, and computer programs were not readily available to verify the results as it happened, indicates that this phenomena has not been tested before ... or at least not on Earthwaves. Then there is the point of weather this test actually showed anything of relevance. It is sunrise on an oasis in the middle of the African desert, and a 5+ quake occurs in the Kermadec Islands. It is dusk on an atoll in the Pacific, and a 5+ quake occurs in Central Italy. It is dusk or sunrise all over the world at any given time, but can it be as simple as chance, that dusk or sunrise could be in the middle of a randomly selected cross on a screen when these quakes occurred ! People have been debating the possibility of a sun or moon trigger for decades, my interpretation based on these results suggests ... they have been looking in the wrong place !
The results of this test could not be clearer. Each event that I determined had correlation with a selected cross, was reported with which terminator threshold was involved, the time and co-ordinates of the event, and an indication to which line of data in the table this event was associated with. There were some discrepancies which I addressed after the test. When one is trying to keep pace with seismic activity, format the relevant data for posting, and try to converse its meaning all through the test period ... discrepancies will happen.
It has always been my understanding (since joining Earthwaves), if a scientist comes forth with a credible hypothesis, that for what ever reason fits a particular picture, it is only right that it be tested by other scientists in the same manner. If there testing produces similar results, then the hypothesis can be dismissed because it has not stood up to scrutiny. If their results fall short of the original test, it can then be considered a working theory. I'm not a scientist, but I ran an experiment in plain sight, over a period of ELEVEN days. I chose this period because my previous attempts had indicated the source was indicative of a cycle, which repeats at specific times. This is still speculation because I need to amass more data during the coming months.
With respect to Roger, I have clearly stated that I know nothing of computer programs. I have no idea what the results mean, I don't know the dates or times chosen , which terminator thresholds made contact where, what period this refers too, and most of all, I am expected to have full confidence in computer results obtained behind closed doors, and calculated from data that was not fully understood 5 days previously. The only real way of testing this phenomena is to repeat the test as many times as is necessary, to get a true picture as to weather this is significant or not.
I have to admit, that I have taken the time to try and understand the meaning of the word "chance". The dictionary quotes chance as being " the incalculable element in existence that renders events unpredictable" so Rogers interpretation is correct in respect to predicting earthquakes. But it continues " an event without an observable cause ", does this second quote have any relevance to what occurred during the test ? I don't know the cause and that's why I am investigating it, but I observed what happened ... contact was made at point A, and an action occurred at point B at the same time !
Lastly ... I remember some time ago, enduring a two paragraph grammar lesson on the correct use of the word " Theory ". It came about in conversation as a " pet peeve " which was particularly annoying to the person I was corresponding with. I thought it trivial at the time because I believed stating theory or hypothesis basically meant the same thing in the eyes of others. But I was the new guy so I complied, and I hope the individual this refers too, has noted I still do ... but I have learned in science, everything has to be correct, or it can be miss-interpreted. So in hopes that I do not develop my own pet peeve, I would like to repeat the following; this test has "nothing" to do with 30 day windows, I did not use this test for prediction purposes, so the test itself can only be analysed in the period stated, in past 11 day periods before new moon (using the same data), or in future 11 day periods before new moon. I have been coached to adopt an honest scientific approach with my project. I could not be more honest with this than do it in real-time ... I ran a fair test, it deserves a fair analysis !
It is not my intention to cause any offence here, the 15th Jan data lines got 9 hits out of the 28 recorded terminator contacts, and 6 of the 15 sun / moon contacts ... statistical analysis could explain the total as chance, but how would it explain this ?
Duffy
RE: A Fair Test 8 / 2 / 2017 - Skywise - 02-08-2017
(02-08-2017, 05:59 PM)Duffy Wrote: I have to admit, that I have taken the time to try and understand the meaning of the word "chance".
Roger's use of the word 'chance' is from a statistics point of view. As well as several other words.
A simple example: coin tossing. It's pretty obvious that the odds of throwing heads is 50%. Now, if you throw 10 tosses (HTTHTHTHHT) with 5 heads and 5 tails, then that's chance.
What you are doing is obviously much more complicated. What roger is trying to do is figure out what the chance odds are of your crosses correlating* with earthquakes. For example, you have 20 crosses, and 15 of them end up with quakes. That's 75% percent of them. But what were the chance odds anyway? If the chance odds ARE 75% then you are doing no better than chance, no better than random guessing, If you do better than chance, then that indicates 'significance'.
Significance also has a meaning in statistics. It's basically a measure of HOW much away from chance something is. This can get quite complicated and is based on things called distributions. Even if you got 90% of your crosses right given a chance of 75%, depending on the distribution curve of a random test, the significance may be high, or low.
By way of comparison, you wouldn't expect 10 coin tosses to ALWAYS have exactly 5 heads and 5 tails. Sometimes you get more of one than the other. 6 heads out of 10 coin tosses is better than chance, but not necessarily significant. It depends on the distribution of the results of many sets of ten coin toss sessions. This is something you can easily try yourself. Do 10 coin flips and record the number of heads. Then do ten more. Do this for at least 10 sets (total 100 coin tosses). Make a chart tallying how many times you get a particular number of heads, that is how many times did you get 0, or 1, or 2, etc, on up to 10. Then graph the results to see the distribution curve. The more sets, the smoother the curve. In this well known example you should get what is called a gaussian or normal distribution. You may have heard it called a bell curve.
*And for more definitions, correlation here means coincide. Science has a saying, though, that "correlation does not equal causation". Just because something correlates doesn't mean one causes the other or has any relationship to it. To drive that point home, check out the Spurious Correlations website - http://www.tylervigen.com/spurious-correlations. But finding correlations is an important first step in scientific progress.
Granted, statistics is not an easily understood subject. But it is such an important tool in the science tool box that its understanding is necessary.
Moving on, it is my observation that there's been a lot of miscommunication. You guys just aren't speaking the same language. I have to admit, I'm having a little trouble totally understanding your method. It happens. Sometimes a thing needs to be explained many different ways until it is comprehended. I've had to do that myself explaining things to others.
I encourage continuing the discussion in spite of these difficulties. I feel that in the end we will all learn something new. Duffy, you will certainly gain a better understanding of science, scientific methods and thinking, and statistics. Roger, I, and others will learn about new ideas in quake prediction (successful or not) and strengthen our knowledge and skills in studying and analyzing quake correlations. Even if your method turns out to be chance and have nothing to do with quakes, it's still important information. Science is built as much on what doesn't work than what does. Perhaps even more so.
I also encourage this discussion if for no other reason than how it is being carried out. It is civilized and constructive. This is how these discussions should be carried out and it is refreshing after years of previously hostile and adversarial discussions in the past.
Brian
RE: A Fair Test 8 / 2 / 2017 - Roger Hunter - 02-08-2017
(02-08-2017, 08:41 PM)Skywise Wrote:(02-08-2017, 05:59 PM)Duffy Wrote: I have to admit, that I have taken the time to try and understand the meaning of the word "chance".
Roger's use of the word 'chance' is from a statistics point of view. As well as several other words.
A simple example: coin tossing. It's pretty obvious that the odds of throwing heads is 50%. Now, if you throw 10 tosses (HTTHTHTHHT) with 5 heads and 5 tails, then that's chance.
What you are doing is obviously much more complicated. What roger is trying to do is figure out what the chance odds are of your crosses correlating* with earthquakes. For example, you have 20 crosses, and 15 of them end up with quakes. That's 75% percent of them. But what were the chance odds anyway? If the chance odds ARE 75% then you are doing no better than chance, no better than random guessing, If you do better than chance, then that indicates 'significance'.
Significance also has a meaning in statistics. It's basically a measure of HOW much away from chance something is. This can get quite complicated and is based on things called distributions. Even if you got 90% of your crosses right given a chance of 75%, depending on the distribution curve of a random test, the significance may be high, or low.
By way of comparison, you wouldn't expect 10 coin tosses to ALWAYS have exactly 5 heads and 5 tails. Sometimes you get more of one than the other. 6 heads out of 10 coin tosses is better than chance, but not necessarily significant. It depends on the distribution of the results of many sets of ten coin toss sessions. This is something you can easily try yourself. Do 10 coin flips and record the number of heads. Then do ten more. Do this for at least 10 sets (total 100 coin tosses). Make a chart tallying how many times you get a particular number of heads, that is how many times did you get 0, or 1, or 2, etc, on up to 10. Then graph the results to see the distribution curve. The more sets, the smoother the curve. In this well known example you should get what is called a gaussian or normal distribution. You may have heard it called a bell curve.
*And for more definitions, correlation here means coincide. Science has a saying, though, that "correlation does not equal causation". Just because something correlates doesn't mean one causes the other or has any relationship to it. To drive that point home, check out the Spurious Correlations website - http://www.tylervigen.com/spurious-correlations. But finding correlations is an important first step in scientific progress.
Granted, statistics is not an easily understood subject. But it is such an important tool in the science tool box that its understanding is necessary.
Moving on, it is my observation that there's been a lot of miscommunication. You guys just aren't speaking the same language. I have to admit, I'm having a little trouble totally understanding your method. It happens. Sometimes a thing needs to be explained many different ways until it is comprehended. I've had to do that myself explaining things to others.
I encourage continuing the discussion in spite of these difficulties. I feel that in the end we will all learn something new. Duffy, you will certainly gain a better understanding of science, scientific methods and thinking, and statistics. Roger, I, and others will learn about new ideas in quake prediction (successful or not) and strengthen our knowledge and skills in studying and analyzing quake correlations. Even if your method turns out to be chance and have nothing to do with quakes, it's still important information. Science is built as much on what doesn't work than what does. Perhaps even more so.
I also encourage this discussion if for no other reason than how it is being carried out. It is civilized and constructive. This is how these discussions should be carried out and it is refreshing after years of previously hostile and adversarial discussions in the past.
Brian
Brian;
Thank you for a well thought out presentation of the situation. I'm not good at it, partly because my two fingers get tired. I will add that any result from coin tossing is chance whether 1 head in 10, 5 in 10 or even 10 in 10.
They all have different probabilities.
One of my motives in continuing this project is the hope of finding something that works but has been overlooked by the standard professional groups as too improbable to bother with.
So far all I've done is show they are right.
Roger
RE: A Fair Test 8 / 2 / 2017 - Roger Hunter - 02-09-2017
(02-08-2017, 09:53 PM)Roger Hunter Wrote:(02-08-2017, 08:41 PM)Skywise Wrote:(02-08-2017, 05:59 PM)Duffy Wrote: I have to admit, that I have taken the time to try and understand the meaning of the word "chance".
Roger's use of the word 'chance' is from a statistics point of view. As well as several other words.
A simple example: coin tossing. It's pretty obvious that the odds of throwing heads is 50%. Now, if you throw 10 tosses (HTTHTHTHHT) with 5 heads and 5 tails, then that's chance.
What you are doing is obviously much more complicated. What roger is trying to do is figure out what the chance odds are of your crosses correlating* with earthquakes. For example, you have 20 crosses, and 15 of them end up with quakes. That's 75% percent of them. But what were the chance odds anyway? If the chance odds ARE 75% then you are doing no better than chance, no better than random guessing, If you do better than chance, then that indicates 'significance'.
Significance also has a meaning in statistics. It's basically a measure of HOW much away from chance something is. This can get quite complicated and is based on things called distributions. Even if you got 90% of your crosses right given a chance of 75%, depending on the distribution curve of a random test, the significance may be high, or low.
By way of comparison, you wouldn't expect 10 coin tosses to ALWAYS have exactly 5 heads and 5 tails. Sometimes you get more of one than the other. 6 heads out of 10 coin tosses is better than chance, but not necessarily significant. It depends on the distribution of the results of many sets of ten coin toss sessions. This is something you can easily try yourself. Do 10 coin flips and record the number of heads. Then do ten more. Do this for at least 10 sets (total 100 coin tosses). Make a chart tallying how many times you get a particular number of heads, that is how many times did you get 0, or 1, or 2, etc, on up to 10. Then graph the results to see the distribution curve. The more sets, the smoother the curve. In this well known example you should get what is called a gaussian or normal distribution. You may have heard it called a bell curve.
*And for more definitions, correlation here means coincide. Science has a saying, though, that "correlation does not equal causation". Just because something correlates doesn't mean one causes the other or has any relationship to it. To drive that point home, check out the Spurious Correlations website - http://www.tylervigen.com/spurious-correlations. But finding correlations is an important first step in scientific progress.
Granted, statistics is not an easily understood subject. But it is such an important tool in the science tool box that its understanding is necessary.
Moving on, it is my observation that there's been a lot of miscommunication. You guys just aren't speaking the same language. I have to admit, I'm having a little trouble totally understanding your method. It happens. Sometimes a thing needs to be explained many different ways until it is comprehended. I've had to do that myself explaining things to others.
I encourage continuing the discussion in spite of these difficulties. I feel that in the end we will all learn something new. Duffy, you will certainly gain a better understanding of science, scientific methods and thinking, and statistics. Roger, I, and others will learn about new ideas in quake prediction (successful or not) and strengthen our knowledge and skills in studying and analyzing quake correlations. Even if your method turns out to be chance and have nothing to do with quakes, it's still important information. Science is built as much on what doesn't work than what does. Perhaps even more so.
I also encourage this discussion if for no other reason than how it is being carried out. It is civilized and constructive. This is how these discussions should be carried out and it is refreshing after years of previously hostile and adversarial discussions in the past.
Brian
Brian;
Thank you for a well thought out presentation of the situation. I'm not good at it, partly because my two fingers get tired. I will add that any result from coin tossing is chance whether 1 head in 10, 5 in 10 or even 10 in 10.
They all have different probabilities.
One of my motives in continuing this project is the hope of finding something that works but has been overlooked by the standard professional groups as too improbable to bother with.
So far all I've done is show they are right.
Roger
Duffy;
More test results.
This time the quake list was all mag 5+ quakes from 1973 to mid January, 2017, 72006 quakes in all
I took a quake every 10.5 days as a random "signal", computed subsolar and sublunar points and then looked for any quakes within 0.5 degrees of the 90 degree line for sun and moon.
The first match within 30 days was counted as a hit.
There were 785 "signals" chosen and 321 hits total found so the chances of getting a hit are 321/785 or 0.409
almost a coin toss situation.
So if you are getting fewer hits than this you can quit looking.
If anyone wants different parameters tried just let me know.
Interestingly, the curve is pretty tight, It only takes 34 more hits to pass my 99% significance test.
Roger
RE: A Fair Test 8 / 2 / 2017 - Skywise - 02-09-2017
(02-09-2017, 05:21 AM)Roger Hunter Wrote: If anyone wants different parameters tried just let me know.
Open the band up from .5 degrees either side to 1 degree. On the face of it, this is doubling the area so one might expect the odds to double. But you know how fickle these things are.
Brian
RE: A Fair Test 8 / 2 / 2017 - Roger Hunter - 02-09-2017
(02-09-2017, 05:39 AM)Skywise Wrote:(02-09-2017, 05:21 AM)Roger Hunter Wrote: If anyone wants different parameters tried just let me know.
Open the band up from .5 degrees either side to 1 degree. On the face of it, this is doubling the area so one might expect the odds to double. But you know how fickle these things are.
Brian
Brian;
First, to correct some errors, the signals were 20.5 days apart and the band was +/- 0.25 degrees wide. Remember there were 2 of them; sun and moon.
This time the bands were +/- 0.5 degrees wide. There were 499 hits, raising the odds to 0.636, a 50% increase.
Roger
RE: A Fair Test 8 / 2 / 2017 - Duffy - 02-09-2017
Hi Brian
Its good to hear from you, thank you for taking the time to give a very thorough explanation, of the meaning of these words. I now understand why Roger and I have been having some problems. Chance in my vocabulary, as an ordinary member of the public has meant insignificant or nothing to gain. So when Roger has been telling me "chance", I have interpreted this as meaningless or why should I be bothering. I will consult your description before I make assumptions again in future. I never did have a knack for tossing coins, but I have gotten pretty good at counting to 10 ... probably the reason why we remained civilised
. Thanks againRoger
I think we both got backed into a corner over this, and didn't realise it. I cannot guarantee this won't happen again, but at least we learn something each time . I'll send you my friends recipe for mulled wine, as way of an apology ... how'd that be ?
Odds 0.636 ... does this mean I keep looking ?
Duffy
RE: A Fair Test 8 / 2 / 2017 - Roger Hunter - 02-09-2017
(02-09-2017, 05:42 PM)Duffy Wrote: Hi BrianDuffy;
Its good to hear from you, thank you for taking the time to give a very thorough explanation, of the meaning of these words. I now understand why Roger and I have been having some problems. Chance in my vocabulary, as an ordinary member of the public has meant insignificant or nothing to gain. So when Roger has been telling me "chance", I have interpreted this as meaningless or why should I be bothering. I will consult your description before I make assumptions again in future. I never did have a knack for tossing coins, but I have gotten pretty good at counting to 10 ... probably the reason why we remained civilised
Roger
I think we both got backed into a corner over this, and didn't realise it. I cannot guarantee this won't happen again, but at least we learn something each time . I'll send you my friends recipe for mulled wine, as way of an apology ... how'd that be ?
Odds 0.636 ... does this mean I keep looking ?
Duffy
Along the lines that Brian laid out I ran my program with different bandwidths to show how the curve looks.
The width I'm quoting is half width in degrees so 0.1 means +/- 0.1 or 0.2 degrees total
With a zero width there were no hits.
With 0.10 there were 169 hits for 0.215 odds on a hit
With 0.25 there were 321 hits for 0.409 odds on a hit
With 0.50 there were 489 hits for 0.636 odds on a hit
With 1.00 there were 655 hits for 0.834 odds on a hit
With 1.50 there were 719 hits for 0.916 odds on a hit
With 2.00 there were 853 hits for 0.959 odds on a hit
With 2.50 there were 769 hits for 0.980 odds on a hit
Now you can compare your results to see if you're doing better than chance. If not, your results do not show any merit.
The program makes this sort of analysis possible. Each run looking at over 72,000 quakes takes about 6 seconds.
Roger
RE: A Fair Test 8 / 2 / 2017 - Duffy - 02-09-2017
Roger;
I am still convinced that this is not the way to go about this, but I will oblige you and try your method first.
If you are using a 90 degree distance from sub-solar, I will of course have to change the dawn/dusk bands to sunrise/sunset for this to be comparable with your method. It will take me 2 or 3 days to compare my results in this manner, so I want to make sure this is acceptable. Also, I use to be able to calculate orbital purturbations of asteroids in years past, but I've never been able to figure odds. I would appreciate a quick 101 on this, or perhaps you can work the figures out for me when I get them !
I would still like to try a real-time test again, if your still willing to provide the data ... I will need it by the 15th if you are.
Duffy
RE: A Fair Test 8 / 2 / 2017 - Duffy - 02-09-2017
(02-09-2017, 10:23 PM)Duffy Wrote: Roger;
I am still convinced that this is not the way to go about this, but I will oblige you and try your method first.
If you are using a 90 degree distance from sub-solar, I will of course have to change the dawn/dusk bands to sunrise/sunset for this to be comparable with your method. It will take me 2 or 3 days to compare my results in this manner, so I want to make sure this is acceptable. Also, I use to be able to calculate orbital purturbations of asteroids in years past, but I've never been able to figure odds. I would appreciate a quick 101 on this, or perhaps you can work the figures out for me when I get them !
I would still like to try a real-time test again, if your still willing to provide the data ... I will need it by the 15th if you are.
Duffy
Sorry Roger,that won't work because it will negate the whole idea of the test. The data from my results is as follows;
Dawn contacts = 5
Dusk contacts = 11
Sunrise = 7
Sunset = 5
You can see from this, dusk was the more influential of the bands. Your program would have to allow for this by 18 degrees. But even this could be difficult because although sunrise/sunset can have a fixed distance, the dawn/dusk bands alter with the sun's North and South motion ... am I understood on this point before I continue ?
Duffy