A comparative study for estimating the parameters of the second order moving average process
Abstract
En
Moving Average process is a representation of a time series written as a finite linear combination of uncorrelated random variables. Our main interest is to compare a classical estimation method; namely Exact Maximum Likelihood Estimation (EMLE) with the Generalized Maximum Entropy (GME) approach for estimating the parameters of the second order moving average processes. In this paper, in applying EMLE we have to find the exact likelihood function through deriving the probability density function of the series. Differentiating the function with respect to the parameters, we can obtain the exact maximum likelihood estimates. On the other hand, the idea of GME is to write the unknown parameters and error terms as the expected value of some proper probability distributions defined over some supports. We carry a simulation study to compare between the presented estimation techniques.
Moving Average process is a representation of a time series written as a finite linear combination of uncorrelated random variables. Our main interest is to compare a classical estimation method; namely Exact Maximum Likelihood Estimation (EMLE) with the Generalized Maximum Entropy (GME) approach for estimating the parameters of the second order moving average processes. In this paper, in applying EMLE we have to find the exact likelihood function through deriving the probability density function of the series. Differentiating the function with respect to the parameters, we can obtain the exact maximum likelihood estimates. On the other hand, the idea of GME is to write the unknown parameters and error terms as the expected value of some proper probability distributions defined over some supports. We carry a simulation study to compare between the presented estimation techniques.
DOI Code:
�
Keywords:
Time series; Moving average; Exact Maximum Likelihood; Generalized maximum entropy
Full Text: PDF
