Quote:
Originally Posted by Robert Tulip
My hypothesis is that lunar-planetary cycles produce non-random effects on rain amounts. Testing this requires a data trawl, but the amount of data I have presented is not enough for proof. I do not have a hypothesis as to whether Uranus might have a bigger effect than other planets such as Jupiter, but am just looking at the data to see. My aim was to define a method able to test objective planetary effects. This is very easy to validate by replacing the Sydney data with data from any other official weather station.
May I say, the heliocentric hypothesis emerged from Copernicus’ observation of the inelegance of epicycles. He subsequently went in search of data to justify the heliocentric view, with only limited initial success due to his assumption of circular rather than elliptical planetary motion. Similarly, I believe this rain effect is elegant, in line with the gravity net hypothesis summarized above, and am in search of data to prove it.
|
Robert, thanks for the reply. I've read your description of your methods a little more carefully, and have a few more comments.
1. Your hypothesis is very vague. "If I look at rainfall data and a lot of lunar/planetary cycles, something will look non-random". In which direction, when and why?
2. If your hypothesis is driven by gravitational/tidal effects, why Uranus? With such a short run of rainfall data, if there's any effect, the Moon should overwhelmingly dominate, with the Sun a distant second, and anything else being utterly negligible.
3. If there's some sort of tidal interaction between the Moon and Uranus, shouldn't we expect to see peaks/troughs on or around days 1 and 15? Instead you see days 5 and 11.
4. Estimating statistical significance by counting standard deviations from the mean is a
very shaky approach for these data. Rainfall data have a very asymmetric distribution, with almost all the readings being near zero, and a scattering of days of very heavy rainfall far out on the right-hand-side of the graph. Hitting a couple of these on the same day of the cycle would give a very high reading purely by chance. I'd try a bootstrap approach, taking a long series of random samples of the data and seeing where your results lie in this empirical distribution. A cheaper and dirtier approach would be to take logs of the rainfall values and look at their distributions, although you'd need to fix the days of zero rainfall (try subtracting 1 from the log of the lowest non-zero reading). This would reduce the influence of the outliers, although you'd still have a very skewed distribution.
5. If you're claiming the effect to be planet-wide, it should be easy to get hold of other rainfall series for validation (assuming you have an effect in the Sydney data, and I'm not convinced you do). If there's nothing else in Australia, there should be plenty of cities in other continents with long runs of rainfall records.
6. I commend you for your generally rational approach to this issue, but please note that you automatically lose 25 credibility points for comparing yourself to Copernicus. I should warn you that if you go on to compare yourself to Galileo or Einstein, you drop off the credibility board altogether.