Forecasting Using Vector Autoregressive Models (VAR)
Forecasting Using Vector Autoregressive Models (VAR)
Forecasting Using Vector Autoregressive Models (VAR)
Volume: 3 Issue: 10
ISSN: 2321-8169
6058 - 6061
______________________________________________________________________________________
Abstract Forecasting data can give a better understanding, control and manage unexpected results better. Vector Autoregressive Models (VAR
Models) have for long been used to find trends in a set of non-discreet values. In this paper, we focus on building VAR models, to determine the
best fit using various tests and the results obtained when we applied VAR models to estimate the future values and trends in soil moisture. Along
with the application of VAR models to predict the soil moisture, we have also additionally applied it to the temperature and mean sea level
pressure forecasting, results of which are presented.
KeywordsVAR; Vector Autoregressive; VWC; Volumetric Water Content; Forecasting;
__________________________________________________*****_________________________________________________
I. INTRODUCTION
Vector Autoregressive (VAR) models can be used to
predict the future values with the help of past values. It could
be applied to any series of values to predict the future
outcome. In a project we carried out, we measured the soil
moisture and estimated the Volumetric Water Content (VWC)
in the soil. The future values of soil moisture were predicted
by using VAR models.
The VAR model has been used applied regularly for
describing the dynamic behavior of economic and financial
time series and for forecasting. Forecasts from VAR models
are quite flexible because they can be made conditional on the
expected future paths of applied variables in the model.
II. STAGES OF VAR MODEL
Following are the steps to be followed to build VAR model.
i. Importing and pre-processing data.
ii. Specifying a model.
a. Specifying Models to set up a model using
vgxset, involving the following
Specification Structures with Known
Parameters.
Specification Structures with No Parameter.
Specification Structures with Selected
Parameter.
b. Determining an Appropriate Number of Lags.
iii. Fitting model to data, which could include,
a.
b.
c.
6058
IJRITCC | October 2015, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
ISSN: 2321-8169
6058 - 6061
______________________________________________________________________________________
A. Adf Test
i.
ii.
[EstSpec,EstStdErrors,LLF,W] = ...
vgxvarx(VARmodell,Yest,[],Ypre);
EstSpec structures are the fitted models.
If the highest lag value is 24 then pre-sampled data
is from 1 to 24, estimated data is from 24 to 90% of complete
data and forecast is remaining 10% of data.
Examine whether fitted models are stable or unstable
by using the function vgxqual.
[isStable,isInvertible] = vgxqual(Spec)
If the value of isStable is 1 then the fitted model
Spec is stable.
VI. ANALYSING AND FORECASTING USING FITTED MODEL
Now that we have different fitted models, we identify
which fitted model is the best fit and use the particular fitted
model for predictions. Analyzing the fitted model involves:
i. Sum of mean squared error.
ii. R-squared test.
A. Sum of Mean Squared Error
_______________________________________________________________________________________
ISSN: 2321-8169
6058 - 6061
______________________________________________________________________________________
_______________________________________________________________________________________
ISSN: 2321-8169
6058 - 6061
______________________________________________________________________________________
will be the predicted value, thus giving a foresight on how the
data being monitored, behave in the future.
[4]
http://bbsdlp.litbang.deptan.go.id/tamp_komoditas.php
[7]
Watson,
M.
(1994).
Vector
Autoregressions
and
Cointegration, in Handbook of Econometrics, Volume IV. R.F.
Engle and D. McFadden, Elsevier Science Ltd., Amsterdam.
6061
IJRITCC | October 2015, Available @ http://www.ijritcc.org
_______________________________________________________________________________________