Anomaly Alert! Text Similarities In Descriptions Of Events Might Depend On Similarities In Astrology

Hello there, I'm Renee O'Shaq, an astrologer and researcher hailing from the vibrant city of Boulder, Colorado, nestled in the heart of the United States. I've stumbled upon a peculiar discovery that I'd like to share with you.

The Cosmic Confluence of Events and Astrology

In my quest for knowledge, I often use Mathematica for my coding projects. Apart from being a mathematical software, it offers a plethora of secondary tools, one of which is an event database. This database provides a start date and a brief description for nearly 8,000 events, a number that may seem insignificant when compared to the grand tapestry of history.

However, this database was not without its constraints. I only included the 7,153 events that began at or after January 1st, 1600 A.D. This decision was made to exclude events with earlier Julian dates. Similarly, I removed numerals from the descriptions to avoid any confusion. For instance, the Battle of 1812 would become the Battle of, without the numerals. This was to ensure that two events described as occurring in the same year but having similar descriptions and astrology charts would not be considered as one.

The Dance of Words and Stars

To measure similarities in the text descriptions of each pair of events, I utilized the scalar dot product of the embeddings from the bi-directional encoder, BERT, from Google. This was a powerful tool that allowed me to quantify the similarities in the charts of the pairs.

For the astrology charts, I used a scalar result of adding the cosines of the differences between the astrology placements for the sun, the moon, Mercury, Venus, Mars, Jupiter, Saturn, Neptune, Uranus, and Pluto, as well as the North and South Lunar Nodes. This resulted in a continuously defined single number that ranged from -12 for chart pairs that were all at 180 degrees across the 12 placements, through 0 for squares, all the way to positive 12 for conjunctions for the 12 placements.

The Independence Test and the Mysterious Dependence

The null hypothesis in this study was that the measures of description similarity and of charge similarity were independent, meaning they were not related. I was looking to test the alternative hypothesis that one measurement is dependent on the other.

To conduct this test, I used Hoofting's independence test, an integral-based statistic that uses ranked values, similar to Spearman's rank. However, calculating the statistic for the full data set was computationally prohibitive, so I had to resort to using a p-hat upper value for the p-value instead.

This approach involved using a multi-color resampling of 1 million random pairs done 100 times. The upper bound for the p-value was then calculated as r + 1 divided by 100 + 1, where r is the number of samplings with a higher p-value than the culturally accepted one for significance of 0.05.

To my surprise, no resampling out of the 100 resulted in such a p-value. In fact, they were infinitesimally small, smaller than the chance of you reaching out your hand and grabbing one particular random electron out of all the electrons in the universe. Thus, the p-hat upper bound for the p-value here is simply 1 divided by 101, which is less than 0.01.

The Cosmic Conclusion

In this study, the null hypothesis of independence turned out to be statistically extremely unlikely. Therefore, the strict converse is implied - that one measurement is dependent on the other. This dance of words and stars, a celestial ballet of astrological charts and historical events, is not as independent as we thought.

For more references and the full calculations, including that for the generation of data, you can use this QR code. Thank you for joining me on this celestial journey.