Abstract:
|
In traditional statistical problems, compete data-sets at all the moments are always available. Statistical methods usually require data with the complete history for the calculation especially when there is a temporal dependence over time. In some applications, the data in the past moments are not saved for different reasons and only the data at the current moment are available. This paper propose a method to estimate the asymptotic variance for the sample mean of temporal dependent data when the past data are not be saved. The asymptotic consistency of the estimate is shown. The memory space needed in the calculation is determined. Simulations were done to check the numerical performance of the proposed method. An application in nuclear engineering is discussed.
|