Normal distance rank tests gave 1.9–4.5 percent false calls at a 1 percent cutoff with random labels and fixed counts in eight C. elegans brain maps

We tested the error rate of a spatial synapse test in eight fixed C. elegans brain maps. The test compares distances within one chemical synapse direction with distances between opposite directions. Distances share synapses. We used 269 neuron-pair records with at least three sites in each direction. Under uniform direction labels with counts and coordinates fixed, a normal distance rank test gave mean false-call rates of 1.90 to 4.49 percent at a 1 percent cutoff. Tie and continuity corrections gave 1.90 to 3.94 percent. We kept the same rank statistic and reassigned whole site labels. We used 216 exact reference laws and 53 sampled laws. Separate validation gave zero to 0.78 percent at the 1 percent cutoff. In the recorded maps, Holm correction within each map left 55 distance-test calls and 23 label-test calls. An exact six-site example showed that distance permutation can also fail without a normal approximation. The package has a reusable test tool, proofs, source hashes, coordinate extracts, reference draws, figures, and repeat instructions. Dependent-distance inference and label permutation are established methods. The new contribution is this connectome benchmark and checked implementation. The label test assumes direction exchangeability at fixed locations. This study does not reproduce the earlier N2U analysis, measure neural function, or test a population growth trend.

Caenorhabditis elegans · connectome · synapse location · reciprocal connections · rank-sum test · dependent distances · random labeling · permutation test · false positive rate · reproducible research