XPost: alt.ufo.reports   
      
   I watched an ep of a repeat of Craig Charles' "UFO Conspiracy" a few   
   days ago. It reviewed the flyby from the "first interstellar asteroid"   
   designated 2017 U1 aka 'Oumuamua "first messenger from afar".   
      
   The usual run-through of the basic facts and some interviews from   
   various experts incl Avi Loeb. Some of the experts underlined the   
   object was strange and accelerated via means unknown after its close   
   approach with the sun in 2018.   
      
   But like all these kinds of programs nothing was decided. Like the   
   rest of the UFO background that emphasises study of individual cases   
   and going through all the juicy details, it ended up "proving" not   
   very much -- even to Charles, a "UFO believer".   
      
   We've looked at 2017 U1 before.  We found that various UFO activity   
   seemed to have a beyond-chance imprint of the movement of Amooamooa   
   around the solar system. My conclusion then was that it seemed   
   unlikely the object was a card-carrying interstellar spaceship -- it's   
   top speed was only seen to be around 50 km/sec -- but that maybe, just   
   maybe, some of the "locals" had gone out to have a look at it because   
   maybe they have a healthy curious about such things (we've prev also   
   seen the imprint of the movements of various Earth-origin probes seem   
   to also appear in UFO activity), and when some of them later turned up   
   on Earth the ups and downs of day to day sightings reports carried an   
   echo of the distance they had travelled out to 2017 U1.   
      
   But the AI's have been playing around with a lot of things since and   
   have come up with some very very special results applicable to   
   interstellar visitors. So we're going to try this on Amooamooa. And   
   the results will be very, very interesting.   
      
   We've seen before with these kinds of studies the idea is to take   
   something like the distance of an object from the earth (generally   
   known as "delta" in my version of English) and see whether it can   
   predict the number of sightings at some later time. It's generally   
   expected that if 2 places are closer together more traffic of all   
   kinds tends to flow between them than if they were further apart. With   
   a distribution like e.g. Zipfian we expect the number of trips between   
   A and B to vary approx as a power of the distance. It's just a matter   
   of finding that power and proving the relationship is predictive and   
   statistically robust. We did that kind of thing with U1 last time.   
      
   But now we are going to tighten the requirements up a whole heap.  Not   
   only must the relationship between delta and UFO sightings of a   
   certain type must determine an appropriate "power law", and the   
   power law must pass 2 basic statistical tests at better than 95%   
   confidence, it must also validate at levels that indicate predictive   
   skill.  Basically the first 1/2 of the data is used to estimate the   
   model, the 2nd 1/2 of the data is used to re-test that model, and the   
   error found in the test must be significant better than just using the   
   average value of Y as a "guess".   
      
   With Amooamooa and other interstellar visitors (there is already a 2nd   
   one to look at and we'll get to that in a later post) we have a   
   special opportunity. We can use the "incoming" 1/2 of the dataset to   
   train the model. And test it on the "outgoing" 1/2 of the   
   dataset. Will the final model be able to skillfully predict with data   
   that is the "opposite" that it saw during training? Will it be able to   
   predict the elephant has a trunk after seeing that little piggy tail?   
      
   Well, of course, I would not be posting this if there was some doubt. :)   
      
   Not only does the search find statistically strong, validated, skillful   
   models for various types of UFO activity, it finds the models show   
   that the closer U1 came to Earth and the sun, the higher certain kinds   
   of activity. As U1 left again that same activity in all cases dropped   
   off again.   
      
   While much of the activity related to different UFO types categorised   
   by a color (i.e. therefore likely some kind of light in the sky,   
   probably seen at night) the same was found for certain types of   
   objects that definitely can not be confused with a light in the sky.   
   In particular, it found that U1's close pass through the solar system   
   was accompanied by a huge upswing in sightings of military aircraft   
   chasing this, that and the other.  Also of note, U1 also seems to be   
   associated with one of my fave types of UFO presently (although I   
   haven't seen any AFAIK) the football-sized "big mother". Like a slew of   
   other types of UFO's, the giant disk seemed to go up in numbers as U1   
   approached the sun, particularly at the times earth was nearest to U1,   
   and they went down again to a "background" level as it left.   
      
   I will set up a sub-dir at  with the   
   various plots and details, but we'll look at a couple, below.   
      
   The AI's used the 4 basic "distance" measures of U1 they dredged up   
   from the JPL database ("Horizons"). The distance between earth and U1   
   ("delta"), the speed of U1 toward/away from the earth ("deldot"), the   
   distance of U1 from the sun ("r"), and the speed of U1 relative to the   
   sun ("rdot").   
      
   The tests applied were so strict no solutions were found for the   
   "rdot" case. For other measures multiple hits against UFO activity of   
   various types -- usu characterised by a color in the "short comment"   
   part a sighting report, or the shape classification assigned by (in   
   this case) the folks at NUFORC. I've also set up a couple data series   
   that are supposed to represent really hard-to-explain sightings that   
   happen "all the time" -- (a) everything except a light in the sky, (b)   
   anything described as "hovering", particularly if overhead, and (c)   
   anything apparently chased by a military jet, light aircraft, black   
   helicopter, or any other "mundane" aircraft.   
      
   Using the "delta" value, that varied for U1 from a large number as it   
   approached in 2014, going to a small number in 2018 as it whipped past   
   the sun, and then increasing again to a largish number in 2021 at the   
   end of the data (in order that the approach and departure had the same   
   number of data-points for its first half/last half training/validating   
   regime) we find the best models that pass all the tests and also   
   validate to high levels of skill were:   
      
   X       Y            Params  R2   
   del     military     4o2     0.31619540   
   del     bright       0o2     0.31403383   
   del     hover        5o2     0.25430144   
   del     Flash        0o2     0.21201243   
   del     dark         3o2     0.17719790   
   del     white        4o2     0.16421241   
   del     Other        6o2xy   0.15286574   
   del     Circle       0o2     0.13187625   
   del     Diamond      2o3     0.11881049   
   del     blue         1o2     0.11832348   
   del     giant        3o1xy   0.09879727   
   del     green        0o1xy   0.08700952   
   del     red          2o3xy   0.07498949   
   del     brown        0o3     0.02535379   
      
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)