Stochastic subspace identification (SSI) is a robust time domain modal identification algorithm. Generally, SSI identifies modal parameters from dynamic responses of structures under zero initial condition, while the effects of large amplitude inputs on the results are neglected. Based on existing studies, this paper presents the effects of large amplitude inputs on SSI and investigates the applicability of this method. First, the theoretical foundation of SSI considering large amplitude inputs is briefly introduced. Then, through a two degrees of freedom system, the relative accuracy of different modes and mode shapes changes with data of samples, and the large amplitude inputs are studied. Finally, a case study of the fluctuation test of the arch rib of the Yangtze River Caiyuanba Bridge is presented to investigate the applicability of the method in identifying large amplitude inputs. The study shows that considering large amplitude inputs can improve the relative accuracy of modal parameters. Compared to natural frequency, the accuracy of mode shape change with data samples shows more randomness. SSI of large amplitude inputs has a high standard for the completeness of modal experiment.