ARMA 모형의 상태공간에 의한 최대 우도 적합에 관한 연구 = Maximum Likelihood Fitting of ARMA Models With State-Space Approach

다국어 초록 (Multilingual Abstract)

The state-space model has been extensively applied to mode18ing data from econemics, medicine, engineering, etc. The model also can be used to the calculation of likelihood function of a stationary autoregressive and moving average(ARMA) time series when a time series data is fitted by maximum likelihood. Here rewriting of ARMA model into state-space model, the form of the likelihood function and Kalman filter, Kalman smoothing techniques are reviewed. Computation of initial state covariance matrix is important when Kalman recursion is applied and its achievement is shown by based on impulse response. The model parameter estimation by maximum likelihood fitting leads to difficult nonlinear optimization technique and here recursive estimation algorithm by EM(expectation maximization) particularly useful instate-space involving unobserved components or irregularly observed data is discussed. And its example is represented.