Analyses are made on the effect of spectral shape on statistical veriability of wave parameters such as heights and periods, the difference between the parameters dedined by the zero-upcrossing and the zero-downcorssing methods, and a spectral parameter governing the wave grouping. Numerically-simulated wave profiles supplemented by field and laboratory data are used.
　The coefficients of cariations of carious parameters are evaluated for four typical wave spectra. A caution for the use of simulation technique is cited. The two definitions of zero-crossing waves yield the same ensemble means of characteristic heights and peirods.
　The difference bewtween them for individual records remains within a range of variadility which is evaluated quantitatively. However, waves exhibiting a large value of newly-defined atiltness parameter produce significant differences in wave periods. The mean run length is closely related to an envelope correlation parameter, which can be calculatyed from the spectrum.