XPost: alt.ufo.reports, alt.paranet   
      
   After some scouring of web resources I've managed to gather together a   
   big list of monster sightings in lakes and oceans.  The longish history   
   of the sightings makes a tantalising study possible -- can we link   
   these data to the movements of planets and asteroids going back   
   centuries? Of particular interest are the movements of Pluto. Normally   
   it's very long orbit -- almost 250y -- means it can't reliably   
   compared with any phenomena that does not go back to at least the 16th cent.   
   With sea monsters, it becomes possible.   
      
   The only small problem we find now is the movement of Pluto itself   
   is not very well known prior to the 18th cent. I had managed to   
   snaffle some data from JPL that purported to show the orbital position   
   of Pluto back to the 17th cent. Just checking today is seems that   
   model has been withdrawn and the present Horizons database only tracks   
   Pluto back to 1800.  But I still have the older version of the data,   
   so let's forge ahead!   
      
   The AI programs have been churning through the problem for several   
   days now, examining many ways of massaging the available data and   
   performing robust analysis thereof, and they seem to be settling down   
   to a less than interesting conclusion the data is just too poor at   
   this point to drawn any detailed conclusions. However they are quite   
   sure there is SOME link between planetary movements (we've also looked   
   at key asteroids going back 100s of years, too, but that I'll leave   
   for a follow-up article) and lake/sea monsters, but exactly which   
   planets have the best link(s) depends on exactly what you do in the   
   analysis.   
      
   An outline of how the calculations proceed is as follows.  We assemble   
   a "total sightings" series based on data going back to the 16th   
   cent. The data-set we're using today consists of sightings for 3 lake   
   monsters -- Caddy, Nessie, and Champ -- plus historical sightings of   
   "sea serpents" in the Atlantic and Pacific. The Champ data consists of   
   2 sources -- a set of roadside signs in Pt Henry, NY, and a 2023 paper   
   by F. Foxon.  The net result is an estimate of the rate of sightings   
   per year going back to 1635.   
      
   (Random sample of 1/2 the data-points:)   
   1635 24.7375 1636 6.7261 1638 13.3755 1639 13.6528 1643 13.5462 1645 14.9221   
   1650 26.4682 1651 19.5114 1652 16.1528 1655 12.2257 1660 2.73077 1661 6.95687   
   1665 23.1477 1666 21.7421 1668 19.1825 1671 25.9682 1674 20.7421 1676 24.7375   
   1677 9.87545 1679 20.5114 1682 23.7375 1684 6.7261 1685 12.3755 1690 22.3319   
   1691 22.3319 1695 10.9564 1699 5.13636 1702 26.2375 1703 9.78571 1704 20.6011   
   1705 20.6011 1707 15.4518 1709 7.2261 1710 3.82051 1711 7.2261 1714 1.58974   
   1719 17.9518 1720 24.7375 1722 22.1011 1727 20.5114 1730 6.36713 1731 9.28571   
   1732 12.4221 1734 13.8154 1735 3.82051 1738 23.6477 1739 23.1477 1740 12.8154   
   1744 17.4518 1748 20.5114 1749 13.8154 1753 16.6825 1754 12.3755 1755 7.86713   
   1757 12.4221 1759 12.1062 1762 13.8154 1764 8.2261 1765 21.7421 1767 6.36713   
   1768 22.1011 1770 24.8785 1771 17.2426 1772 8.28571 1775 14.862 1776 4.08974   
   1777 12.5462 1780 17.6825 1782 3.08974 1783 16.2426 1785 15.1528 1786 16.9518   
   1787 13.8154 1788 23.1477 1794 17.6825 1796 9.28571 1799 13.6062 1801 16.7426   
   1802 10.8755 1806 15.0462 1807 11.3755 1808 12.3154 1811 22.1011 1812 12.2257   
   1814 11.3154 1818 13.4221 1819 14.862 1820 6.86713 1824 15.362 1825 5.32051   
   1827 18.0114 1829 6.86713 1830 2.23077 1831 8.2261 1832 25.9682 1833 24.3785   
   1834 13.9221 1835 21.7421 1836 9.45687 1837 9.51648 1838 15.4518 1839 5.7261   
   1842 11.9564 1843 10.7857 1844 11.3154 1848 2.73077 1851 13.6062 1852 13.0165   
   1854 9.72566 1859 1.5 1861 10.5165 1863 14.6528 1864 15.9221 1866 10.5165   
   1867 24.7375 1869 24.3785 1870 25.3785 1871 10.5165 1874 25.9682 1876 18.0928   
   1877 22.5114 1878 4 1879 11.6364 1881 18.5928 1882 16.9221 1883 23.6477   
   1885 12.7857 1887 10.7857 1889 16.6528 1890 23.7421 1893 17.7426 1897 14.0462   
   1902 17.7426 1903 7.95687 1904 19.6011 1906 12.5462 1909 6.95687 1912 12.3154   
   1915 8.45687 1916 11.2857 1917 8.95687 1919 1 1923 16.6528 1924 8.86713   
   1925 20.7421 1929 2 1930 4.82051 1931 13.0165 1933 96.7308 1934 133.922   
   1938 33.4569 1939 19.3671 1941 3.23077 1947 21.8755 1949 10.7857 1950 17.4221   
   1951 20.0118 1952 20.3205 1955 9.08974 1956 6.32051 1957 16.8755 1963 20.5897   
   1964 24.1364 1965 23.7857 1966 35.7857 1969 19.3205 1970 11.7261 1971 24.4569   
   1972 14.5 1974 20.6364 1975 20.4569 1976 29 1978 29.8205 1980 24.8205   
   1981 63.2308 1982 35.5 1983 44.7308 1984 37 1986 23.3671 1987 22.2308   
   1991 7 1992 13.1364 1993 7.23077 1995 19.7426 1996 24.2857 1998 17.6364   
   2001 14.2857 2002 12.3671 2003 22.2426 2005 6.23077 2006 9.13636 2007 4.23077   
   2009 6.13636 2012 12.5165 2015 23.6528 2016 11.8205 2018 17.8205 2019 32.6528   
   2020 18.1364   
      
   What we're going to do is create a series of "validated models"   
   against the positions of the major planets (and Pluto!) going back as   
   far as the JPL data goes. By "validated" we mean part of the data will   
   be used to estimate the model -- how the sightings vary depending on   
   the avg annual distance of each planet from the Earth -- and then that   
   will be checked to ensure the rest of the data is explained just as   
   well. To up the pressure on the models they will have to meet a series   
   of stringent requirements -- not only must they validate (as above)   
   but they must be statistically relevant as well.  They must pass at   
   least 1 usual stats test (e.g. rank test or T-test on the \beta of a   
   time series model), they must all be "power models" i.e.  relate the   
   sightings rate with some power of the distance of the planet from   
   Earth, and the model must show the closer the planet the greater the   
   number of expected sightings. I've also added in one other thing the   
   AI's sometimes get away with. The model must USE the data from each   
   century of the data-set rather than ignore it. Sometimes a small number   
   of data-points (e.g. 10%) are too far from the predictions a model   
   makes and the procedure usually is to go back and ignore those points   
   in order to make the model "more certain". I wont care about overall   
   certaintly  this time -- I just want to make sure every century of the   
   dataset contributes in some way to the estimation of the model.   
      
   To make things even harder on the AI's, we also specify the last 1/4   
   of the data is all that can be used for training the model; however the   
   remaining 75% must be found to fit the same model without any mucking around!   
      
   And, so, after some days of twiddling, including going back several times   
   to re-combine the known data into the estimate of "total monster   
   sightings for the year", we finally have the following list of results:   
      
      
   Validation	Filter	Trans	Planet/object   
   0.649872        1.5 		jupbary   
   0.650533        1.5 		satbary   
   0.652327        1.5	x 	uranus   
   0.653412        1.5	x 	mars   
   0.657618        1.5	x 	neptune   
      
   [continued in next message]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)