This paper pays a special attention to the issue that there is a significant scatter of the stress-strain responses of a nuclear engineering material, 1Cr18Ni9Ti stainless steel pipe-weld metal. Statistical investigation is made to the cyclic stress amplitudes of this material. Three considerations are given. They consist of the total fit, the consistency with fatigue physics and the safety in practice of the seven commonly used statistical distributions, namely Weibull (twoand three-parameter), normal, lognormal, extreme minimum value, extreme maximum value and exponential. Results reveal that the data follow meanwhile the seven distributions but the local effects of the distributions yield a significant difference. Any of the normal, lognormal, extreme minimum value and extreme maximum value distributions might be an appropriate assumed distribution for characterizing the data. The normal and extreme minimum models are excellent. Other distributions do not fit the data as they violate two or three of the mentioned considerations.