Newsgroups: alt.ufo.reports,alt.paranet.ufo
From: MrPostingRobot@kymhorsell.com
Subject: another trunk of the elephant moment

[uploaded 2 times; last 23/09/2024]

A visitor was asking me a couple days ago how sure could anyone be of
what some machine learning algorithm picks out from a mess of numbers
no-one really can comprehend.

I was frank. No-one can really be sure of anything most of the time.
But having some program pick out things from a mess of material is
usually better than guessing. If it's programmed right (sometimes the
biggest "if") at least the pick is objective and considered as many
side issues as machinegly possible. Many studies have looked
favourably on some simple program versus a panel of experts in a
subject area. The studies would not have been worth publishing unless
their conclusion involved the panel of experts being wrong to very wrong
most of the time. :)

I've emailed off the following example to the person as a typical
thing that programs come up with, that may not seem very convincing
initially, but when you look into it further you find there is definitely
something to what they concluded.

I've been looking at data for small earthquakes across the world.  
It seems they come and go and correlate robustly will all manner of
things. Presently there is no real reason why this should be. But we
may have our suspicions.

As part of this the AIs at Quant Central have been looking at how
almost undetectable earthquakes, particularly in central California,
seem to correlate with the comings and goings of key asteroids.  We've
looked at *that* kind of thing in the past, and the same patterns
repeat. Certain asteroids seem to predict robustly these small quakes.
In physics you would be thinking of the close approach of some
asteroid -- maybe it doesn't get closer than 100x further than the
moon, so it's not "close" to most people's way of thinking -- seemed
to be followed 24 hrs later by a small quake in central California
then information travelled from that asteroid to central California.
Information movement requires a particle of some kind.  OK. You can
see where this will be going. ;)

So one of the recent tasks under the relevant heading was to look at
which asteroids in the local database (only about 2000 at present
with data on a day-to-day basis from ~2000) highly correlate with
small quakes in which areas and depths of California.  That's a lot of
number crunching and takes poor old baby a couple days to get through.

But a couple days back it spat out the top 10 leading contenders it
had found.  Interestingly, it DIDN'T find "most" asteroids seemed to
predict California quakes, but a surprising 10% seemed to be
implicated.  On a day by day basis the top contender worked out to be
asteroid 50620680.

Not much is known about this beast. Databases don't know it's size,
rotation period, or much else apart from its orbit. JPL lists the
"osculating elements" thus:

  EPOCH=  2459455.5 ! 2021-Aug-29.00 (TDB)         Residual RMS= .16429
   EC= .225725439512327    QR= .9925291709347452   TP= 2459518.8511386891
   OM= 128.6459221282158   W=  270.1860779675507   IN= 8.987489544889289

The QR shows its perihelion is just inside the earth's orbit.  So
it's definitely "near earth". And INclination is only 9 deg -- way
above the ecliptic but not extreme compared to many others previous
studies at Quant Central have come up with for mysterious happenings
(inclinations of 60 deg or even higher come up a lot).

The program did all its checking and determined that the distance
between earth and asteroid 50620680 explained more than 1/2 the small
earthquakes all over California at a depth around 12.5 km. There
aren't many detectable that far down. The kind of quakes we're talking
about have a mostly -ve mag -- usually characterised as too small to
be detected by humans at the surface. But they can bump a needled on
the relevant instrument OK.

I got a list of the top 10 asteroids via morning email from the AIs
and decided to take a look at the raw data to see how convincing it
was to me. This is what they had found for asteroid 50620680:

Date		   Dist (AU)        #quakes     model est #quakes
2020.161               2.12838            1     0.977301
2021.678              0.124046            3      2.52471
2021.721              0.131789            5      2.47418**
2021.891              0.154204            4      2.34776**
2022.377               1.13567            1      1.20535
2022.697               2.06852            1     0.986655
2022.902               2.23616            1     0.961314
2023.366               2.00064            2     0.997709**
2023.467               2.10734            1     0.980549

The above data has been edited to remove the boring bits.  It's daily
data from 2020 to end 2023. On most of those days either there is 0
quakes at 12.5 km anywhere in California, or only 1 detected.  But we
can see there is a small cluster of events in mid 2021 where there
were 3, 5 and 4 on consecutive days. And at the time asteroid
50620680 was near its minimum distance from earth at only .12 AU out
-- about 45x the distance to the moon.

According to the stats this stands out like a sore thumb as
significant in the 2 ways the programs normally test -- a T-test and
Rank test on the relevant time series regression.

But we can also see another data-point I left in the sample -- in
early 2023 there was a day with 2 small quakes, but at THAT time the
asteroid was 2 AU out -- almost to its max distance.

So maybe we doubt this relationship really holds water.

But the AI's are using more than just the stats tests. All they have
learned in the past 5y is going into the mix somewhere.  And it's
their balanced judgement -- after taking a vote between themselves --
that this particular dataset is significant and they suggest is the
best they can find in the sample of 2000 asteroids presently at hand.

So if we are pretending to do science, then let's try to push the
model over!  This highlights what has been a bug-bear with me with
this kind of data -- almost everything you look at somehow correlates
with UFO activity at least "somewhat". You don't seem to be able to
pin anything down to just one location, one type, a small number of
any kind of choices.  It's one big fuzzy mess. Researchers in the area
talk about rabbit holes.  I've written about "gateway phenomena"
elsewhere. (As in gateway drug; not fuzzy circles that transport you
to other places or dimensions).

So I did some checking. This particular asteroid only comes close
every now and then. Being "near earth" means it's possible that "close
approach" events can be very rare -- e.g. if the orbit of an asteroid
is very similar to earth's then the times they are very much in sync
can be few and far between. If something has a period of 7 days, and
another has a period of 8 days then they only synch every 56 days. 
Kinda thing.

So looking back in the asteroid data -- that in the database goes back
to the 1950s -- I found the prev close approach was in 2005. The
reason that had not been checked until now -- the AI's were just using
a common period they had for all data and that was between 2020 and
2023. So I went to the right program and downloaded the quake data for
2005 +- 2 years.  Q: Would it show the same kind of pattern?   ...

Well, you know me by now. Why am I posting this if it didn't knock my
socks off.  A: Pretty much exactly.

Here's the data for the prev close approach:

2003.544                  2.09658            1
2004.284                  2.63454            1
2004.730                  1.63994            1
2004.872                  1.00843            1
2005.358                  0.43156            1  <-closest approach
2005.372                 0.449716            1
2005.443                  0.51176            3 <- similar pattern
2005.445                 0.512959            3 <- similar pattern
2005.448                 0.514069            2 <- similar pattern
2005.454                 0.516027            1
2005.456                 0.516876            1
2005.497                 0.519523            1
2005.500                 0.519061            1
2005.514                 0.515705            1
2005.533                 0.508471            1
2005.552                 0.499034            1
2005.645                 0.462097            1
2005.664                 0.467662            1
2005.779                 0.690324            1
2005.836                 0.900738            1

It could not be much more similar. The original data showed closest
approach associated with 3 consec days above normal.  The 2005 data
shows closest approach a couple of days before the same kind of pattern.

Man. Another elephant trunk situation. :)

--
Making successful predictions is the gold standard of science. If a
theory successfully predicts phenomena that are later observed, one
can be confident that the theory captures something essential about
the real world system.
-- Andrew Dessler, testimony to US Senate, 21 Jan 2014

From stardust to pale blue dot: Carbon's interstellar journey to Earth
new.nsf.gov, 14 Apr 2021
Most of the carbon on Earth was likely delivered from the interstellar
medium, the material that exists in space between stars in a galaxy.

[Martians Bad Fer Bidness!]
US Scientist Warns Excessive Alien Talk Could Create Societal Problems
CNN-News18, 09 Sep 2024 05:41Z
Recently, a former US intelligence officer claimed that aliens are hiding on
Earth. These talks have led to people believing ...

Building block of life found in sample from asteroid Ryugu
Space.com, 21 Mar 2023
The discovery of the nucleobase uracil is a big step forward for astrobiology.
One of the four nucleobases of RNA has been discovered in samples retrieved
from the asteroid Ryugu, providing the strongest evidence yet that the
organic building blocks for life on Earth came from space.

Aliens May Be Doomed by Inescapable Climate Change, Study Finds
ScienceAlert, 23 Sep 2024 03:37Z
Earth's average global temperatures have been steadily increasing since the
Industrial Revolution.

Senate defense committee to hold UFO hearing
The Hill on MSN.com, 17 Sep 2024 23:52Z
The Senate Armed Services Committee is looking to hold a UFO hearing after
the November elections, Sen. Kirsten Gillibrand's ...

  Truth is out there: Congress to get update on UFOs from Pentagon as
  Gillibrand presses for data
  New York Post, 15 Sep 2024 19:30Z
  Congress will soon be getting an update from the Pentagon on UFOs, according
  to New York Sen. Kirsten Gillibrand -- who says ...

  New UFO hearings to take place in Senate this month after new sighting
  MSN, 15 Sep 2024 18:34Z
  Another round of public UFO hearings are in store for Congress within
  mere weeks, according to a senior member of the Senate's Armed Services
  Committee.

  New UFO hearings to take place in Senate this month after new sighting
  Daily Mail on MSN.com, 15 Sep 2024 17:17Z
  The new Senate inquiry comes as many American civilians have reported their
  own UFO sightings - including a 'huge' UFO just ...

Incredible pics of `first UFO selfie' resurface showing glowing spherical
craft with yellow `landing lights'
The Sun, 13 Sep 2024 13:44Z
A BRAZILLIAN alien researcher has snapped what many believe is the
first-ever UFO selfie. The snap was taken by UFO ...