Convergence Tester
This tool has been withdrawn. An internal review on 23 August 2026 found that its permutation null model did not do what the interface said it did. The convergence statistic it permuted depends only on the set of values, not their order, so shuffling could not change it — every p-value the tool produced was 1.0, on every dataset, and the combined Fisher test was degenerate for the same reason.
Any result obtained from this tool should not be used or cited, including the methods paragraph it generated for pasting into manuscripts. If you used it in work that is unpublished, the analysis needs redoing; if in work already submitted or published, please contact us and we will help you correct it.
The statistic itself needs redesigning rather than the code being patched, so this page will stay as it is until that work is done and independently checked.
A second and separate correction, 23 August 2026. This tool shipped preloaded example datasets labelled “Preloaded Dataset” under the instruction “Enter ethnographic observations for each culture”, attributed to the Martu, the Karuk, the Malian savanna, and to Nepali, Māori and San/Khoisan traditions. Those values were not drawn from our research data, and we cannot trace them to any source. An initial version of this notice said they came from a manuscript held pending cultural review. That was wrong, and we are correcting it here rather than quietly.
They do not match our own fire-management dataset and in places contradict it: that analysis records low fire intensity for all three traditions, while the preloaded example showed moderate for one; it records rotation medians of 1.0, 9.5 and 1.0 years against the example’s 3, 5 and 2; and it states that patch-size distributions are essentially unreported in the literature for these traditions, while the example supplied a hectare figure for a culture for which our own paper says no such figure exists. The values appear to have been written as filler to demonstrate the interface, and were then presented to readers as observations about named communities. That should not have happened, and no preloaded cultural dataset will appear in any rebuild of this tool.