Microsoft Word Thesis Gent (1). doc
Chapter VI. Methodology
Download 1.76 Mb. Pdf ko'rish
|
DPTX 2010 2 0 330292 0 110731
6
Chapter VI. Methodology 6.1 VAR methodology The vector autoregressions (VARs) are a macro-econometric models provided in 1980 by Sims. Based on this provision, a univariate autoregression can be defined as a linear model which uses one variable which is described by its lagged values. Extending this model, the vector autoregression is a multi-variable linear equation where variables are explained by their individual lagged values, as well as the values of other past lasting variables as n-1. This macroeconomic model has assisted to understand various time series and has a very easy statistical background to be used. The vector autoregressions (VARs) have been observed as very helpful in describing data, prediction, and analysis of policies. This simplistic framework engages linear equations to observe the relations of variables that are endogenous. The VAR model accounts for all of variables as symmetrical and does not include a theory to show the variables as dependent or independent. They are based on previous data and analyzed afterwards. The VAR model treats all variables as endogenous and has a system with multi- equation. It is a model that has one equation for every variable as dependent variable. Furthermore, in these equations there are included lagged values of all variables of the model as dependent variable including the independent variable. These equations have the same form, because for VAR model there are no contemporaneous variables that are used as explanatory variables. For example, in a VAR model with variable as y the equations of the model would be as below: y t = a 1 y t−1 +... + a k y t−k + e t y Furthermore, in this model the endogenous variable serve also as explanatory variables in lagged form, while the amount of lags is determined at the estimation stage. In order to see that which lag length suit my estimation better, I run the information criterion for lag length. According to information criterion Akaike (AIC) the goodness of fit that relatively suits my model better is number of lags between three or four. 46 The VAR model has several properties’ that are: • It is a reduced form as the on right-hand side variables there are no contemporaneous variables. • The variables in the VAR model depend on each other, as they are all considered as endogenous variables • The shocks in the VAR model are unobserved structural shocks. Thus as the data in VAR estimation are real data, the estimates obtained are of combined shock and are denoted as e to be differentiated from the structural shocks. With these shocks there are constructed impulse response analyses. • The VAR model is usually used to forecast, while it is not used for structural analysis and policy evaluation. Download 1.76 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling