Abstract
The goal of this paper is to test the interest rate models for China's repo market to understand the behavior of China short rate. I obtain the functional nonparametric estimates of drift and diffusion terms. Because the interest rate process may not be stationary and it has very different properties in two subperiods, I consider their local time estimation. Furthermore, I find that the density of the process is bimodal and it follows Vasicek model within each of the two subperiods. So I assume that the interest rate process also relies on a state variable in addition to short rate. I use a two-regime model to fit the data and study its properties, such as the probabilities that the process stays in one regime and transition probabilities from one regime to another.