We present methods and results to reduce fake burst events induced by nonstationary noises. To reduce these fake events, we systematically surveyed monitor signals recorded with a main (or gravitational-wave) signal of a gravitational-wave detector so as to watch the detector. Our survey was to check whether or not there was a coincidence between the main and monitor signals when we found a burst event from the main signal. If there was a coincidence, we rejected this event as a fake event induced by nonstationary noises, regarding the main signal as being dominated by nonstationary noises. As a result, we succeeded to reject about 90% of the burst events of which the SNR values were larger than 10 as fake events, with an accidental probability of about 5% to reject burst-gravitational-wave candidates.