Scribd
Upload
37 views

Workshop Sym

The document discusses different types of data used in time series analysis including time series data, cross-sectional data, and panel data. It also discusses stationary and non-stationary data, with non-stationary data having a variable mean and variance over time. The autoregressive distributed lag (ARDL) model is presented as a technique used when some variables are stationary and others are non-stationary at level 1, but not at level 2. The steps to build and validate an ARDL model are outlined.

Uploaded by

Danish Hameed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
37 views

Workshop Sym

The document discusses different types of data used in time series analysis including time series data, cross-sectional data, and panel data. It also discusses stationary and non-stationary data, with non-stationary data having a variable mean and variance over time. The autoregressive distributed lag (ARDL) model is presented as a technique used when some variables are stationary and others are non-stationary at level 1, but not at level 2. The steps to build and validate an ARDL model are outlined.

Uploaded by

Danish Hameed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 29

TIME SERIES DATA

 Quantities that represent or trace the


values taken by a variable over a
period such as a month, quarter or
year.
 Time series data occurs wherever the
same measurements are recorded on
a regular basis.
e.g.
 Closing prices of stocks
 Country’s unemployment rate
CROSS-SECTIONAL DATA
 It is a type of data collected by
observing many subjects (such as
individuals, firms, countries or
regions) at the same point of time, or
without regard to differences in time.

e.g.
 GDP of European countries in 2010.
 Unemployment rate of Asian
countries in 2014.
PANEL DATA
 Data where multiple cases (people,
firms, countries etc) were observed
at two or more time periods.

 It combines both TIME SERIES and


CROSS-SECTIONAL data.

e.g.
 Income and expenditure of different
employees in different time (years)
EXAMPLE OF PANEL DATA
PERSON YEAR INCOME AGE SEX
1 2011 1300 27 1
1 2012 1600 28 1
1 2013 2000 29 1
2 2011 2000 38 2
2 2012 2500 39 2
3 2013 2800 40 2
NON-STATIONARY DATA
 Non-stationary data, as a rule, are
unpredictable and cannot be
modeled or forecasted.

 The results obtained by using non-


stationary time series may be
spurious in that they may indicate a
relationship between two variables
where one does not exist.
NON-STATIONARY DATA
 It has a variable variance and a mean
that does not remain near, or returns
to a long-run mean over time
 e.g.
SEE THE GRAPH ON NEXT SLIDE
STATIONARY DATA
 The stationary process reverts
around a constant long-term mean
and has a constant variance
independent of time.

 It has a constant mean, variances and


covariances over time.

 e.g. SEE THE GRAPH ON NEXT SLIDE.


TYPES OF NON-STATIONARY PROCESSES
 Pure Random Walk (Yt = Yt-1 + εt )

 Random Walk with Drift


(Yt = α + Yt-1 + εt )

 Deterministic Trend (Yt = α + βt + εt )

 Random Walk with Drift and


Deterministic Trend
(Yt = α + Yt-1 + βt + εt )
TREND
AND
DIFFERENCE
STATIONARY
AUTOREGRESSIVE DISTRIBUTED
LAG MODEL
(ARDL)
Conditions under which ARDL is used

 Time Series Data

 Dependent & Independent Variables

 Lag effect of time on both DV & IV

 Some variables are stationary at I(0)


and some are at I(1)
ARDL (continue…..)
 Some variables are I(1) and some
may be I(0), but none is I(2)

 GDPt = αt + βt(Exp) + βt(Cons) + ε t

 First we find optimum levels of lags.

 Through Akaike Criterion(AIC) and


Schwarz Criterion(SIC)
ARDL (continue…..)
 Criteria for optimum lags :

 Lower the value of AIC & SIC, better


the ARDL model

 In order to apply above, formulate


different models with different lags.

 Quick -> Estimate Equation -> Enter


six lags.
ARDL (continue…..)
 D(GDP) C D(GDP(-1)) D(GDP(-2))
…………….. D(CONS(-6)) including
other three variables GDP(-1) EXP(-1)
 CONS(-1)

 Last three variables are included to


check if they have long-run
relationship
ARDL (continue…..)
 Now, check SERIAL CORRELATION

View -> Residual Diag. -> Serial


 Correlation-LM Test -> enter lag=2

 Ho : Model has no Serial Correlation

 It is also desirable for the best model.


ARDL (continue…..)
 Now, check STABILITY
 View -> Stability Diag. -> Recursive
Estimates -> click CUSUM Test

 ARDL estimates must lie in between


two red lines.
ARDL (continue…..)
 To check LONG-RUN ASSOCIATION :
(Bound Test)

View -> Coeff. Diag. -> Wald Test

 Ho = Variables have no long-run


relationship, means don’t move
together.

 Enter : C(8) = C(9) = C(10) = 0


ARDL (continue…..)
 The result brings ‘F’ statistics value.

 It is to be compared with Pesaran


Value with at 0.05.

 Pesaran Table gives Lower and Upper


Bound.

 If ‘F’ stats is > Upper Bound, Ho is


rejected
ARDL (continue…..)
 Hence, all variables have long-run
relationship (move together).

 Now, let’s develop the LONG-RUN


MODEL
ARDL (continue…..)
 Quick -> Estimate Equation

 Derive the Residual, copy and paste it


and name it as ECT (Error Correction
Term), then press OK.

 Now, develop the model with lag 2


including ECT(-1), then press OK.
ARDL (continue…..)
 The results with ECT(-1) will show
“SPEED-UP ADJUSTMENT” means
speed to adjust towards long-run
equilibrium.
 Criteria : the value of ECT(-1) should
be ‘-ve’ and ‘significant’.
 If ECT(-1) = 0.649, means the whole
system can get back to long-run
equilibrium at 64.9%
ARDL
 After including ECT(-1) we need to
check Serial Correlation and Stability
again.

If they are ok, the model is



considered to be the BEST.

 If not, exclude one variable and


check it again.
ARDL
 Finally, Short-Run Causality is also to
be checked which is not desirable for
ARDL Model.

 Therefore, ARDL Model should have :


-No Serial Correlation
-Stable
-Long-Run Equilibrium
-No Short-Run Causality from IV
towards DV.

You might also like