More Ways to Distinguish Mickey Mouse from Interesting Examples

If we do careful and correct measures of significance, we will see night and day differences between the expected and interesting cases. To evaluate a table from the Torah, for example: the best methods involve some combination of:

• mathematical calculation
• computer comparisons of ELS's in the Torah to ELS's in thousands of "monkey texts" (each such text is a purposely mixed-up version of the Torah; for example, the text has the same words within each verse as the real Torah, but in a different order).
• computer comparisons of ELS's in the Torah to ELS's in other real Hebrew texts, modern and ancient.

The tests tend to give answers in reasonable agreement with each other. For example, if a properly done calculation results in a probability of 1 in 10,000; we find that we have to search several thousand texts to find one as good as the Torah. And we conclude that we are looking at an interesting example.

But here, as in all of the steps, we must be careful: we need to be able to properly evaluate whether a text is "as good as the Torah". We must follow the same rules as used for the Torah, but we must not require the same exact ELS's unless those are truly the only choices. Rather we accept a table of similar relevance and rarity, arranged at least as compactly as the table found in the Torah.

Doing this fairly, without bias, once again requires careful analysis. The critical step is to give each comparison text the same level of choice as was available in the original Torah find. And the Mickey Mouse examples often are reported with outrageous odds because this step - if done at all - was done using completely incorrect procedures: such as using scenario 1 for the Torah find but forcing scenario 2 onto the comparison texts.

There can also be very subtle issues of differences between real texts and monkey texts; or dependencies, such as between the words being searched. But with enough repetitions and varied types of tests, and enough varied categories of codes, these problems are recognized and circumvented.

Back One Level
Back to Home Page