In this paper, we focus on the convergence of stochastic differential equations with Markovian switching and 1∕2-Hölder continuous diffusion coefficients. We give the convergence between numerical solutions and explicit solutions at a rate of 1∕logn by the Euler–Maruyama method. Parameter estimations for CIR model with Markovian switching are obtained by the quadratic variation method and composite likelihood method.