Forecasting
Topic:
Times Series Econometrics
By:
Zaheer Khan
Kamroon Taj
To:
Dr. M. Afzal
Time Series Econometrics
Linear Analysis
Non Linear Analysis
Linear Analysis Model
Mean E(Yt) = µ
Variance var(Yt) = E(Yt- µ)² = σ²
Covariance γk = E[(Yt-µ)(Yt+k-µ)]
Stationarity
Ho=ρ=1
Data is non-stationary
This hypothesis is going to be tested
Correlogram
Included observations: 88
LNPCE 0.817576 0.124656 6.558656 0.0000
LNPDI 0.301080 0.094193 3.196406 0.0020
LNPROF 1.387688 0.262312 5.290212 0.0000
LNDVD 1.406317 1.100136 1.278312 0.2048
C 666.7647 187.3353 3.559205 0.0006
AR(1) 0.789442 0.069469 11.36394 0.0000
R-squared 0.998912 Mean dependent var 3877.017
Adjusted R-squared 0.998845 S.D. dependent var 624.4732
S.E. of regression 21.22632 Akaike info criterion 9.014833
Sum squared resid 36495.09 Schwarz criterion 9.184895
Log likelihood -386.1452 F-statistic 14870.78
Durbin-Watson stat 2.007510 Prob(F-statistic) 0.000000
Inverted AR Roots .79
When White test apply along with AR1 this
will reduce the problem of cross products
but still it shows the problem of
autocorrelation
Graphicall Analysis
Correlogram
To check the randomness of the data Q-
statistics is apply which plot the
autocorrelation
Correlogram
Interpretation
Unit Root Test
The graphicall analysis shows that the data is not stationary,
for this the Unit Root Test(ADF) is apply, on the integrated
order 1.
Stationarity of data is the necessary condition of the time
series analysis.
Unit Root Test
The unit root test shows that the data is stationary at the first
order integration
In the observed data after applying the unit root test the data
become stationary
Graphical Analysis
After the test apply there is the visual view of the stationarity
of data.
Which is obtained by plotting the residual of the observation
Residual