XPost: alt.paranet.ufo   
      
   EXECUTIVE SUMMARY:   
   - A perhaps chance -ve corr between N Hem and S Hem UFO sightings allows   
     us to "guesstimate" a rough upper bound on the number of objects that   
     may be responsible for UFO sightings.   
   - UFO sightings density wrt US military bases allows us to categorize   
     types of UFO.   
      
   The coverage of the S Hem in some of the big ufo sighting datasets is   
   a bit spotty. Looking at the NUFORC data upto mid 2020 it seems only   
   about 1/10th of sightings reported to AUS- and NZ-based organisations   
   seem to be making it into N Hem databases.   
      
   While sightings in at least AUS appear to be very much on the decline   
   (and there seem to be interesting reasons for this that will be   
   covered in a later post) the data I managed to scrape, beg, borrow or   
   the other thing suggest an even more interesting possibility.  That   
   the number of "objects" related to UFO sightings may be finite and   
   relatively small.   
      
   This finding comes from the observation that N Hem and S Hem sightings   
   are negatively correlated. I.e. the data shows when there are more   
   sightings in the N Hem there are fewer in the S Hem, and vice versa.   
      
   Here is the basic data for sightings in the S Hem from the New   
   Zealand, NSW, Qld, Vic and WA ufo orgs:   
      
   Year.mm	#sightings   
   1978.79	1	1988.04	2	1993.04	2	1993.62	2	1992.12	1	   
   1992.88	1	1997.71	5	1997.96	1	1997.54	1	1997.46	3	   
   1998.04	6	1998.12	4	1998.29	12	1998.38	26	1998.46	10	   
   1998.54	18	1998.62	40	1997.62	1	1997.79	1	1998.71	18	   
   1998.79	26	2000.62	16	2000.71	10	1999.79	6	1998.88	24	   
   1998.96	22	1999.04	24	1999.12	20	1999.21	18	1999.29	48	   
   1999.38	30	1999.46	37	1999.54	25	1999.62	20	1999.71	12	   
   1999.88	7	1999.96	26	2000.04	23	2000.12	15	2000.21	22	   
   2000.29	12	2000.38	23	2000.46	9	2000.54	14	2000.79	4	   
   2001.04	10	2001.12	18	2001.21	18	2001.29	24	2001.38	10	   
   2001.46	14	2001.54	18	2000.88	10	2000.96	14	2002.46	22	   
   2002.62	12	2002.71	12	2002.79	14	2002.88	2	2003.04	4	   
   2003.12	4	2003.21	6	2003.29	12	2003.38	12	2003.46	14	   
   2002.54	28	2001.62	10	2001.71	12	2001.79	14	2001.96	4	   
   2002.04	2	2002.12	12	2002.21	10	2002.29	14	2002.38	12	   
   2004.21	12	2003.62	22	2003.71	4	2003.79	6	2003.88	2	   
   2003.96	6	2004.04	10	2004.12	18	2003.54	14	2005.21	10	   
   2004.29	12	2004.38	12	2004.46	8	2004.54	6	2004.62	20	   
   2004.71	12	2004.79	4	2004.88	14	2004.96	18	2005.04	2	   
   2005.12	2	2005.46	10	2005.54	4	2005.62	16	2005.71	6	   
   2005.79	12	2005.88	2	2005.96	20	2006.04	4	2006.12	12	   
   2006.29	16	2006.46	8	2006.54	14	2006.62	8	2006.71	16	   
   2006.79	10	2006.88	10	2006.96	10	2007.04	10	2007.12	8	   
   2006.21	8	2007.21	14	2006.38	38	2007.29	16	2007.46	2	   
   2007.54	2	2007.62	14	2007.71	14	2007.79	8	2007.88	24	   
   2007.38	14	2009.79	4	2008.88	10	2007.96	4	2008.04	18	   
   2008.12	12	2008.21	10	2008.29	8	2008.38	14	2008.46	12	   
   2008.54	12	2008.62	6	2008.71	12	2008.79	8	2008.96	16	   
   2009.04	8	2009.21	24	2009.29	16	2009.38	6	2009.46	4	   
   2009.54	6	2009.62	22	2009.71	8	2009.96	6	2010.04	8	   
   2010.12	2	2010.21	6	2010.29	14	2009.88	12	2010.62	2	   
   2010.71	6	2010.88	2	2011.12	2	2011.21	2	2012.21	2	   
   2011.29	8	2011.38	4	2011.46	2	2011.79	2	2012.12	2	   
   2012.29	16	2012.38	8	2012.46	8	2012.54	4	2012.62	8	   
   2015.62	4	2015.54	2	2015.79	2	2016.38	4	2016.54	2	   
   2016.62	2	2016.79	4	2016.88	4	2017.04	4	2017.12	4	   
   2017.21	18	2017.38	20	2017.46	6	2017.62	10	2017.79	4	   
   2016.96	6	2016.12	2	2016.29	2	2018.04	2	2018.29	2	   
      
   And when you compare the NHem (e.g. from NUFORC) and SHem data you get   
   the regression model:   
      
   Durbin-Watson d = 1.388633   
   d < dL (1.683684):  Positive auto-corr at 5%   
   Auto corr detected: estimated rho = 0.312167   
   y = -0.00623419*x + 10.0599   
   beta in -0.00623419 +- 0.00440448  90% CI   
   alpha in 10.0599 +- 1.43569   
   P(beta<0.000000) = 0.989607   
   r2 = 0.0549856   
   calculated Spearman corr = -0.286898   
   Critical Spearman = 0.306000 2-sided at 5%   
      
   The "critical number" is the \beta. It says for each 100 sightings in   
   the N Hem there are about 0.62 (+-.44) *fewer* sightings in the S Hem   
   in the month-by-month numbers.   
      
   The sightings data is VERY noisy as with anything to do with UFO's --   
   most of them are mistakes and something else, after all. ;) So the R2   
   is very tiny at 0.05 or so. Nevertheless the stats tests say the   
   probability the \beta is -ve is around 99%. Even the Spearman test   
   shows the ordering of the data by SHem numbers is so similar to the   
   reverse of the NHem numbers it is unlikely to just be down to luck.   
      
   So, given the NHem and SHem sightings data appear to have a -ve   
   correlation we can then reason this might be because there are a   
   constant number of "objects" and we might try to estimate if that is   
   true what that number might be.   
      
   Knowing "how many" might give us a very rough idea of how many   
   "people" might be involved in creating these sightings. :)   
      
   It's estimated about 4500 people developed and built the 1600-odd   
   '747s, for example.   
      
   To estimate "how many" we take the basic model:   
      
   y = -0.00623419*x + 10.0599   
      
   where y is the number of sightings in the S Hem in a given month, x is   
   the number of sightings in the N Hem for that month, and imagine we   
   can PUSH all the S Hem sightings up to the N Hem by setting y to 0 and   
   solve for x:   
      
   x = 10.0599/0.00623419 = 1613.67.   
      
   Wow! A fairly smallish number and REMARKABLY similar to the number of   
   jumbo jets manufactured.   
      
   BUT.   
      
   There are some other wrinkles we had better take into account.   
      
   The N Hem sightings are mostly down to the USA; the S Hem sightings (I   
   have here) mostly from AUS. And the 2 countries are a bit different.   
   The populations, for one thing.   
      
   If we assume the "ufo density" per month and per km2 over the earth is   
   approx constant we might predict sighting numbers will increased for   
   countries with larger areas and/or larger populations.   
      
   If we take those factors into account the S Hem sightings numbers need   
   to be multiplied by around 13.5 to be comparable with the N Hem.   
      
   I.e. the total number of objects responsible for all the sightings   
   might be "only" 1614*13.5 = 21789.   
      
   This is still not too outlandish. The "lights in the sky" seem to   
   behave in many ways more like unmanned drones than 747's.  It's   
   estimated that (human ;) countries around the world operate a total   
   around 1.3 mn drone aircraft.   
      
   ~ ~ ~   
      
   I set some "homework" last time to estimate the density of USA UFO   
   sightings as they relate to the location of military bases.  I posted   
   last time that the "total" of UFO sightings in the US seems to follow   
   a "square root" distribution -- density of sightings rises according   
   to the sqrt of the distance to nearby military bases.  The square root   
   relationship was a mild surprise but could be related to UFO's in bulk   
   trying to avoid possible contact with ground-to-air missiles that --   
   no coincidence -- have a travel time roughly proportional to the sqrt   
   of the distance to the target (roughly continuous acceleration after   
   firing).   
      
   I was wondering if anyone might be able to break down the same data   
   according to UFO "shape" as given in e.g. the NUFORC database.   
      
   Here are my results, along with some interpretations ;):   
      
   Type		a	R2   
   Cone		0.009	0.98167		avoiding sighting by any base   
   					no matter how distant   
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)