This paper compares two methods for measuring statistical significance in Torah codes research: text randomizations versus permutation tests. The author argues that text randomizations are more appropriate for evaluating the overall existence of Torah codes, while permutation tests measure specific features.
The paper reviews previous Torah codes research, including the seminal 1994 paper by Witztum, Rips and Rosenberg (WRR) that examined equidistant letter sequences (ELS) in the Book of Genesis. WRR used a permutation test to evaluate the significance of compact configurations of ELSs spelling related words.
The author proposes using text randomizations as a more direct approach to evaluate significance. This involves measuring the same word pair configurations in many randomized versions of Genesis created by permuting words within verses. The significance is determined by comparing the results in Genesis to the randomized texts.
The paper applies text randomizations to re-analyze several previous "communities experiments":
For all three experiments, the significance levels measured by text randomizations were substantially stronger than those from the original permutation tests. The author concludes the correlations between ELSs of personalities' names and associated place names are statistically significant in Genesis, regardless of which dataset is used.
The author argues these results show text randomizations are the appropriate significance test for Torah codes experiments exploring statistical evidence for their existence, rather than specific characteristics. The dramatic differences in results between the two methods highlight the importance of using text randomizations to properly evaluate the overall phenomenon.