The statistics in this article have been revised, and the forecast tool it describes has been updated to match.
Two problems were found. The analysis was run on all 62 performances the venue has held, but the resident complaint map did not exist for the first twelve of them; those dates were carrying zero submissions for reasons unrelated to sound. And a joining error credited some submissions to two consecutive shows at once, double-counting 26 of them across five dates.
Correcting both strengthened the reported relationships rather than weakening them: wind direction moved from +0.42 to +0.43, and low-frequency content from +0.40 to +0.51. The complaint-centroid bearing moved from 19 to 18.8 degrees, and the genre medians and comparable-night figures changed accordingly.
The article also reported a held-out correlation of +0.274 without stating how it was calculated. That figure could not be reproduced. It has been replaced with +0.42, computed by a method now described in the text.
