In the first report authors clarified characters of the extremal distribution of a response to ocean waves measured by a full scale experiment. In the paper it was shown that the response was described by a double exponential distribution, which had another character than that of a Rayleigh distribution. To find the theoretical base of the distribution, authors study on various factors which may have some effect on it. The following characters are found by our study. The band-width parameter of the power spectrum for the response has very little effect on the distribution. So it is always approximated by that of a Rayleigh distribution. This means that the distribution does not change by kind of response. There are various ways in statistical treatment of the measured record. Strictly speaking maximum values in the record include those with negative value, but usually they are excluded. Frequently only the largest value for one 0-cross cycle is counted and other maximums for the cycle is excluded. It is clarified by our examination that the distribution is not affected by the difference of the above treatment. The distribution for a record measured under condition with very small value of response has an abnormal wide scatter range ocasionally. It is required to exclude such data in order to estimate the distribution with reliability. The scatter range of the distribution becomes a little wider by drifting of surface and ship conditions during measurement. Regarding the above characters, the distribution is reestimated from the data shown in the first report. In addition to the data, the distribution of various responses is estimated from records of long-term experiment on a pure car carrier and a bulk carrier. By the analysis, it is certified that the extremal distribution is approximated by that of a Rayleigh distribution.