Integration-plot test for peri-stimulus time histograms in human motor units.

A statistical test is proposed for peri-stimulus time histograms in human motor units for the case where a test stimulus is delivered at a constant interval after a previous discharge. This mathematically described test included the notion of the multiple comparison and thus achieved higher sensitivity than the previously proposed method. With regard to data acquisition, the interval from a sham stimulus to the next discharge was acquired as a control, and the total number of samples was set to be four times as large as that in the test situation to reduce the statistical scattering noise. A newly defined statistical object, the integration plot (timewise accumulation of the test histogram without control subtraction) was used for this statistical test. The integration plot had less noise than the cumulative sum (CUSUM) plot (1/square root(2) in theory) and thus represented the neural effect. To compare this integration-plot test with that of the CUSUM, a simulation experiment that compared two sample histograms (one of which had a faint structural change from 20 ms) was conducted. As a result, the present test succeeded in detecting the onset of the change point earlier (23 ms on average) than the CUSUM test (27 ms on average), and the detection probability was also higher (9 out of 10) than the CUSUM (6 out of 10). It was therefore confirmed experimentally that the present statistical test had higher sensitivity than that of the CUSUM proposed previously.