Many schemes have been proposed to perform a model-independent constraint on cosmological dynamics, such as nonparametric dark energy equation of state (EoS) omega(z) or the deceleration parameter q(z). These methods usually contain derivative proces
ses with respect to observational data with noise. However, it still remains remarkably uncertain when one estimates the numerical differentiation, especially the corresponding truncation errors. In this work, we introduce a global numerical differentiation method, first formulated by Reinsch(1967), which is smoothed by cubic spline functions. The optimal solution is obtained by minimizing the functional Phi(f). To investigate the potential of the algorithm further, we apply it to the estimation of the transition redshift z_{t} with simulated expansion rate E(z) based on observational Hubble parameter data(OHD). An effective method to determine the free parameter S appearing in Reinsch Splines is provided.
