Patterns in the Archive

What this is — and what it is not

This page is a set of statistical relationships computed fresh from the archive’s own data every time it grows — not artificial intelligence, and not a substitute for the individual case-by-case analysis this site is built on. Nothing here is trained on anything. Nothing here infers, predicts, or decides. Every number below is a count, an average, or a distance comparison that anyone with the same public files and a spreadsheet would reproduce exactly, and the script that produces them is a few hundred lines of ordinary arithmetic.

There is one real sense in which this page gets better as the archive grows, and it is worth stating precisely rather than dressing up: these are sample statistics over a sample. As case files, locations and sighting records are added, the subgroup sample sizes grow with them, the margins of error narrow, and a relationship that currently cannot be told apart from random noise may become clearly real — or may turn out to have been noise all along. Some of what you read below will change. That is the point of recomputing it, and the changes get recorded on the Network Log the same way every other growth in the archive does.

Two things this page deliberately does not do. It does not report p-values: with a few hundred case files and subgroups chosen after looking at the data, a p-value would claim more rigour than the method can support. Instead each comparison shows its real sample size, its effect size, and — where a null model exists — exactly where the observed value falls inside it. And it does not treat any pattern as evidence for an explanation. A statistical association is an observation about this dataset; it is not a finding about the phenomenon. Where a mundane reading of a result exists, it is given directly alongside the result, not tucked into a footnote.