It is essential in radiobiology to fit a survival-radiation dose curve from experimental data, which was usually done on a computer by the least square method. However, using different objective functions, the curve parameters from the same data may be different. The statistics inference in this paper shows that the objective function ,http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=42648724&type=,http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=42648726&type=,http://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=42648725&type=, is more suitable for survival curve fitting than those used by previous authors.,Three Survival models are used to verify the theoretical inference and a good fitting is achieved. Moreover, this objective function will be applicable to other kinds of fitting work for treating experimental data ranging over several orders.