Cell controls leave reliability matching uncertain after signal adjustment in one selected MICrONS mouse

We tested whether cells with similar natural-movie response reliability connect more often in one MICrONS mouse. The sample had 275,681 candidate pairs, 7,529 connections, 144 source cells and 10,591 target cells. We fitted a linear probability model with both source and target cell intercepts and geometry splines. The product of within-group standardized reliability had a slope of 0.0959 percentage points, with a nominal 95 percent two-way cell-cluster interval of 0.0017 to 0.1901. With measured and modeled signal correlation controls, the slope was 0.0904, with interval -0.0036 to 0.1845. This interval includes zero. The evidence does not establish a reliability-matched subnetwork. The main model placed 31.3 percent of fitted values below zero. Selected cells, one animal, model error and fixed reliability estimates limit inference. New code, aggregate results, source hashes and a separate retrieval script support reproduction. Raw and row-level source records are outside this bundle.

MICrONS · response reliability · cell fixed effects · connectomics · reliability matching · null result