Optimal sampling frequency for high frequency data using a finite mixture model

Advances in IT technology have led to the feasibility of high frequency sampling for stock price data in financial time series. A standard approach to obtaining a sampling cycle is to calculate nn by minimizing the mean squared error (MSE) which is not appropriate for a nonlinear time series mixture and does not account for the number of parameters included in a model and targeted statistical power. The objective of this article is to show two methods for the calculation of optimal sampling frequency under the framework of a finite mixture model. First we investigate how sampling frequency can be obtained through a modified likelihood ratio test to obtain a designated statistical power. Second, we propose a new method based upon the minimization of AICc with a penalty for the number of parameters. Numerical studies show that it performs better than other methods.