To clarify and verify the ultralow frequency (ULF) seismomagnetic phenomena, we have performed statistical studies on the geomagnetic data observed at the Kakioka (KAK) station, Japan, during 2001-2010. We investigated the energy of ULF geomagnetic signals of the frequency around 0.01-Hz using wavelet transform analysis. To minimize the influences of artificial noises and global geomagnetic perturbations, we used only the geomagnetic data observed at nighttime (LT 2:30 A.M. to 4:00 A.M.) and utilized observations from a remote station, Kanoya, as a reference. Statistical results of superposed epoch analysis have indicated that ULF magnetic anomalies are more likely to appear before sizable earthquake events (Es->-108) rather than after them, especially 6-15-days before the events. Further statistical investigations show clearly that the ULF geomagnetic anomalies at KAK station are more sensitive to larger and closer events. Finally, we have evaluated the precursory information of ULF geomagnetic signals for local sizable earthquakes using Molchan's error diagram. The probability gain is around 1.6 against a Poisson model. The above results have indicated that the ULF seismomagnetic phenomena at KAK clearly contain precursory information and have a possibility of improving the forecasting of large earthquakes.