XPost: alt.paranet.ufo
EXECUTIVE SUMMARY:
- Seaice in key areas around Alaska seem to predict daily UFO
sightings some time later.
- The biggest association comes from the Bering Straight and some
large areas along longitude 170W.
- A small region appears in the Mackenzie Bay nr the AK/YT border.
- Given other data associated with UFO sightings nr the Bering Sea and
Aleutians we might suspect UFOs regularly travel via air or water
along 170W perhaps on their way to visit the US or other points S.
- Basic images and plots at .
We've seen in a previous post that certain key regions have an
affinity or predictive power for UFO sightings.
We might interpret a region where e.g. thickness of sea ice is
negatively correlated with daily UFO sightings across the US 30 days
later as somehow "encouraging" more UFO sightings when sea ice
disappears.
The distribution of these particular regions has a pattern we might
expect for other "folkloric" information going back decades. It turns
out not just any region that is subject to seasonal seaice can closely
predict UFO activity.
We've seen that seaice across the Arctic Ocean contains key regions
where the correlation between seaice and UFO sightings seems to go
very high (~90%). From the very rough map of the region posted
previously we could almost see a path from N Russia across the N
America where the data suggested a UFO track might be highly probable.
Here we'll zoom in at least part of the "destination" region in
question and look at the satellite radar data gathered around Alaska.
As before the s/w has gathered together daily NOAA radar maps for the
region from 2010 to the present. For each smaller 100x100 pixel region
of each image a time series is created and correlated against daily
NUFORC sightings data. The s/w finds an optimum lag (days) that
maximize the R2 statistic -- the fraction of day-to-day variations in
UFO sightings that seem to exactly correspond with similar day-to-day
variations in radar returns (aka "sea ice proxy") from each area
across the Alaska region.
Big surprise, a lot of the region shows no association whatever.
Seaice across most of the Pacific off Alaska has no predictive power
for UFO sightings across N Am.
But equally no surprise, some key regions "light up" as if they were
key flight or submarine routes between (wherever) and N Am,
particularly the USA.
The biggest "lit up" area is along longitude 170W starting at the
Bering Straight.
It's as if UFO's normally travel along this route and are inhibited
(maybe ever so slightly :) by the presence of icebergs or seasonal seaice.
The best areas have about an 80% R2. E.g. the stat model for one of
the "best" predictive areas along 170W looks like:
(Log transform enabled).
(58 day lag).
(No serial corr detected).
y = 12.0597*exp(-0.128055*x)
Doubling Rate -5.41
beta in -0.128055 +- 0.0179935 90% CI (42 df)
alpha in 2.48987 +- 0.0203901
P(beta<0.000000) = 1.000000
r2 = 0.77331859
Binned data:
Bin label av radar return av daily UFO predicted UFO
sightings sightings
2010209 -0.642513 12 13.0939* (+1sd over obs)
2015141 0.799166 10.6667 10.8866
2018061 -0.121636 11.3333 12.249
2018194 -2.03485 15 15.6495
2018273 -1.08003 13.2381 13.8484
2019056 0.477173 11.75 11.3449
2019078 1.00353 10.4 10.6054
2019101 2.25662 9.5 9.03311
2019132 0.693845 13 11.0344**(-2sd)
2019228 -1.40318 13.45 14.4335
2019232 -1.2387 13 14.1327*(+1sd)
2019288 -1.75659 17.4286 15.1017*(-1)
2019318 1.31776 9 10.1871*(+1)
2019358 1.64419 9.375 9.77005
2020022 -0.266284 13.125 12.478
2020033 1.3734 10.5 10.1148
2020048 -0.710915 14.2 13.2091
2020053 0.000560869 14.25 12.0588**(-2)
2020112 0.928496 11.45 10.7078
2020128 0.0698588 12.4 11.9523
2020138 1.08781 10.7778 10.4915
2020148 1.56027 9.16667 9.87561
2020153 1.16237 10.1333 10.3918
2020157 1.70302 9.5 9.69672
2020161 0.750138 11.2593 10.9551
2020188 1.27131 9.7619 10.2479
2020198 0.624474 11 11.1329
2020201 -0.361818 13.72 12.6316*(-1)
2020218 -1.34287 13 14.3225*(+1)
2020228 -1.85943 13.9 15.3019*(+1)
2020237 -1.18838 12.0526 14.0419*(+1)
2020250 -1.68671 17 14.9672*(-1)
2020254 -0.815058 13.7391 13.3864
2020274 0.425204 12.5667 11.4206*(-1)
2020293 -0.618811 12.5 13.0542
2020298 -1.29014 16 14.2261*(-1)
2020305 -1.52538 13.8889 14.6612
2020315 0.851102 9.80645 10.8144*(+1)
2020325 -0.429844 13.25 12.7421
2020335 0.150114 11.7 11.8301
2020345 0.538486 10.186 11.2561*(+1)
2020347 -0.504824 13.7368 12.865
2020355 1.47428 9.9 9.98496
2020365 -0.0701027 12.4286 12.1684
This is a binned time-series regression. Datapoints are assigned to
bins that contain "similar" points. The average X and Y is computed
for each bin and those numbers passed to a time-series sensitive
regression. The final statistics are from that regression. In this
case the time-series regression found no problems with serial/auto
correlation so the output is "the same" as an OLS for the same dataset
(with a log transform on the Y data).
The X data is a "normalized" version of the radar return data. Bright
areas on the radar image represent points that have a large
return. These might be flat areas of land or ocean, or sea ice. By
tuning the model to ignore certain signal strengths some of the
clutter (e.g. land or missing satellite data) can be ignored. The
normalization involves translating the return strength into "Z scores"
that have an average of 0 and standard deviation of 1. I.e. a value
of -1 represents a signal around the 16% weakest level detected in the
dataset and +1 represents a signal around the 16% strongest in the dataset.
On that basis the model finds a reduction in signal of 5.4 is
[continued in next message]
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
