Remi/Anonymous
1) You said ,my %hit is 60% and % miss is 13%.It comes to total 73%.What about remaining 23%?
2) Counting clusters of quakes (ie total 165 per for 2010),for averaging, is unfair
3) I have a very simple logic
If my predictions are not good than random dates,any body,you, can do the same thing.If we both give dates till ,say, next six months,than the chances of success or failure are equal for both of us.
so, by this blog I invite ,one and all ,who wish to test the random dates, should post their dates on this blog by January 15th 2011.
The rules are
1) quakes predicted should be 6+
2) time window period is + or - one day ,ie 3 day window
3)If time is given ,it should be in UTC
4) Magnitude window is + or -0.3
5) If time is given, the time window period will be + or -36 Hrs (ie 3 day window)
One,which you claim statisticaly,need to be proved practicaly.
so if 10 persons give radom dates at least 8 should have 96% hits,as you claim
regards
Amit
12 comments:
Dear AMIT,
I simply point out that you have not done any serious statistics.
What you present as statistics are not serious.
It is you who should convince others not me.
I think Roger another gentleman has been saying the same.
Regards,
REMI
I will give you 10 random dates for JANUARY 2011 They are the following...
3,6,9,12,15,18,21,24,27,30
AMIT, I did say Null Hypothesis, but......Interpreting earthquake predictions is difficult. So is evaluating whether predictions
work. To use a statistical hypothesis test, something must be assumed to be
random, and its probability distribution under the null hypothesis must be known.
Many studies model seismicity as random under the null hypothesis. That approach
has serious drawbacks, and details of the stochastic model, such as spatial heterogeneity,
independence or exchangeability, matter for testing. Most null hypotheses used in tests ignore the empirical clustering of earthquakes. Some try to remove
clustering with ad hoc adjustments as a prelude to probability calculations. It is implausible that the resulting data represent a realization of a Poisson process, as is often assumed. The standard approach to testing—hold the predictions fixed while seismicity varies randomly according to some stochastic model—does not take into account that in practice, the predictions would be different if the seismicity were different. The result is that simple-minded schemes, such as the “automatic alarm strategy,” can succeed well beyond chance in hypothesis tests. This is not because the predictions are good: It is because the tests are bad.
The remaining 23% you have not given us prediction. You are only predicting as I pointed out a small fraction of earthquakes as your model does not CAPTURE ALL quakes. The problem is the stats you use is not fair in the first instance. Poisson process for aftershocks is better but there is not widely accepted and persuasive description of aftershock sequences in space time magnitude domain.
Anonymous
What is the use of giving an open window? covering all the dates of the January 2011 ,as window period is + or - one day.
This is not done.
Regards
Amit
Anonymous
1) I have given 45 days ie 135 days with window period.Hence, 225 days left in a year.
This is not 23%
2) I have predicted 77 quakes out of 165,hence 88 left
This is also not 23%
Can you please explain how 23% came?
Amit
Here I repost your data...
Hi Amit, I am glad you will post real statistics and not Micky Mouse hit and misses when the slot boundaries are fixed and not arbitrary around the hit date when it suits us. I am not sure however the physics of the retro moves with change in momentum is understood. In fact it maybe wrong. Thats why I ask you to explain better what momentum changes are there in basic calculations at least. It is important.
Thanks again
Remi
1:00 AM
Blogger AMIT said...
hi
During year 2010 ,ie 1st January to 22nd December the results are as follows
1)6+ quakes occurred--165
2) number of date predicted-45
3) Window period + or - 1 day ,ie 3 day window
4)% of time of the year under window period--37%
5)37 % of total quakes is --61
6) Actual 6+ quakes in window period is ----77
Thus ,it is well above average quakes
Amit
There are 122 3 day slots in a year.
165 quakes occurred in the year 2010.
You predicted 45. (27% of total occurred), very low percentage which tell us that the link must be incomplete and there is a lot more to earthquakes than this link you aspire for a start. ...OK...lets proceed....
Lets see how good you were when you actually made a prediction even if the predictions are a small part of the destructive quakes.....
You are predicting 37% of the window period. So lets see how good you were in this window....
You should have predicted 62 earthquakes. (although we had 77)
you had 37 predicted....8 misses..
% hits = 37/62=60% success not 82%.
% misses=13%.
so 60% success
13% misses
27% not able to predict.
TOTAL = 100%
You said 23% in you main post above ....but it is 27% I did not.
Hi Remi
At least he is giving us a time to watch out. what science and those pretty confusing phrases like /null hypoth.. has done?at least he has the courage to work on it in different way, remember Edison. and Do not measure everything with pure science scale :)
Hi Remi
At least he is giving us a time to watch out. what science and those pretty confusing phrases like /null hypoth.. has done?at least he has the courage to work on it in different way, remember Edison. and Do not measure everything with pure science scale :)
khandan
Thank you very much for your kind support
science,seismology and plate tectonics can not explain some of aspects of earthquakes forget predicting..like
1)why there are quakes at places other than plate boundaries ?
2) Why there are several major quakes on a given date ,all over the globe and on various plates?
3)Why on some given dates major quakes and volcanic eruptions occurs simultaneously?
The problem is we are treating effect (plates) as the cause (of quake)
After so much research and millions of Dollars put in if we/science, predict like this-
-There will be a major quake at certain place either tomorrow or 500 years later- what is the use of such prediction?
DO we have to remain alert 24*7 till 500 years?
I do not understand what is the harm in keeping vigil on some 25 odd dates in a year at prone places?
Any way thanks again
regards
Amit
Hello there! I could have sworn I've been to this site before but after reading through some of the post I realized it's new to me.
Nonetheless, I'm definitely delighted I found it
and I'll be book-marking and checking back often!
Post a Comment