بعض طرائق التشخيص والتقدير في الانحدار الذاتي الخطي وغير الخطي ذي الرتب الدنيا == Some Method of Identification And Estimation For Linear And Nonlinear Autoregressive Models With Lower Order
Author name:
ﻓﺭﺍﺱ ﺍﺤﻤﺩ ﻤﺤﻤﺩ ﺍﻟﻤﻬﻨﺎ
Supervisor name:
عبد المجيد حمزة الناصر
General topic:
Administration and Economics
Specific topic:
Statistics
Degree:
Doctorate
University:
University of Baghdad - Faculty Of Administration And Economics - Department Of Statistics
Language:
Arabic
University location:
Baghdad
First pages:
07T3943 - p.pdf
Abstract:
Time Series Analysis face Two important Problems, take Part in the base of Statistical Work, Which are : Identification and Estimation Problems Specially in Stationary limits. Thus, the identification Process Was based in most of the researches on the Linear Models, Where had used many Methods, Some of them by graphics, and some Other by identification's Criterion's for instance : AIC and BIC Criterions and Others. On the Other hand, the estimation Process stands also on the parametric Methods only, for many years ago. But, the Complexity and improvement of the Phenomena's, and the Progress made in Computers, In both hardware and Software, and the Progress happened in the Numerical Analysis, Courage's the researchers to make their Papers and articles in advanced topics deal with Problems Which could face the identification and Estimation, either the difficulty of large sample size , or by the nonlinear Patterns which could suit some models from others. Furthermore, some Phenomena's may change its behavior at some point (Period) of time, so, it may have two different Patterns (Models) or more. Therefore, at the same time the researcher introduce the Classical identification (AIC , BIC) ,also introduce the new ones ( Table - C and FPE that depended on Kernel function. Where the researcher Make a Suggestion Kernel function in identification. Also we concern with Estimation parameter of Linear and nonlinear Model AR(P) , EXPAR(P) , SETAR(L,K,K,…,K). Here , the researcher also introduce a suggestion of Estimation method grouping the advantages of each of Parametric Estimation Like (Burg and Fisher) and nonparametric like (Spline) together. Also the researcher offers and there proposition about using MGCV Criterion for Estimation ( Spline Method) so in comparison. Also researcher introduce a Suggesting , using the initial Value for Smoothing parameter in Estimation Method (Spline). The researcher used Simulation experiments, by using ( Monte - Carlo) Method, where applied with different Sample Size and different Parameters, Where first : The identification suggesting shows an efficient Criterion especially With the model AR and SETAR , Second : for the estimation suggesting which shows an efficient criterion compared with all the models applied in the simulation experiments
