Scribd
Upload
195 views

Time Series Analysis Forecasting

The document discusses various time series models for forecasting, including ARIMA, VAR, ARCH, and GARCH. It provides the basic formulations for univariate AR, MA, and ARMA models. For ARIMA models, it explains that the p parameter indicates the number of autoregressive lags, d indicates the order of differencing, and q indicates the number of moving average lags. The document also outlines the Box-Jenkins methodology for identifying, estimating, and diagnostically checking ARIMA models.

Uploaded by

Uttam Biswas
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
195 views

Time Series Analysis Forecasting

The document discusses various time series models for forecasting, including ARIMA, VAR, ARCH, and GARCH. It provides the basic formulations for univariate AR, MA, and ARMA models. For ARIMA models, it explains that the p parameter indicates the number of autoregressive lags, d indicates the order of differencing, and q indicates the number of moving average lags. The document also outlines the Box-Jenkins methodology for identifying, estimating, and diagnostically checking ARIMA models.

Uploaded by

Uttam Biswas
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 18

Time Series Analysis

Forecasting
Forecasting using time series
• ARIMA
methodology:univariate/multivariate,
single/multiple equation(s)
• Vector Auto Regression: Multivariate,
multiple equations model
– Granger causality
• ARCH/GARCH:univariate forecasting of
volatility/variance
Atheoretic model, past data predicts future
Univariate single equation models

• AR A R (1 ) : y t     1 y t1  e t
A R ( 2 ) : y t     1 y t1   2 y t 2  e t
A R ( p ) : y t     1 y t 1   2 y t  2  . . .   p y t  p  e t
• MA M A (1 )  y t     0 e t   1e t1
M A ( 2 )  y t     0 e t   1e t1   2 e t 2
M A ( q )  y t     0 e t   1 e t 1   2 e t  2  . . .   q e t  q
• ARMA A R M A ( 1 ,1 )  y t     1 y t1   0 e t   1 e t 1
A R M A ( 2 ,1 )  y t     1 y t1   2 y t 2   0 e t   1 e t 1
A R M A (1 , 2 )  y t     1 y t1   0 e t   1 e t 1   2 e t 2
ARIMA(p,d,q)
• p indicates the no. of autoregressive lags
• d indicates the order of integration of variable
• q indicates no. of moving moving average
lags
• ARIMA(2,0,1)
• ARIMA(2,1,2
• Box-Jenkins methodology helps us decide p
and q.
Box Jenkins methodology
• Identification of p, d and q
– Correlogram
– Partial correlogram
• Estimation
• Diagnostic checking
– Accept model if residuals are stationary, if not
try another model
• Forecasting
Identification using ACF& PCF
• The ACF plots the correlations between yt and yt-k
against the lag k = 1, 2, 3, …: identifies possible MA
terms
• The PACF plots the coefficients in a regression of yt on
yt-1, yt-2, …yt-k against k = 1, 2, 3, …: identifies
possible AR terms
Type of Typical patttern of ACF Typical patttern of
model PACF
AR(p) Decays exponentially Significant spikes
through lags p
MA(q) Significant spikes through
lags p Decays exponentially

ARMA(p.q) Decays exponentially Decays exponentially


Estimation
• Large time series data set requires
• First ensure the series is stationary. If not
make it DS or TS.
• Estimation can be using OLS in most cases
• Regress the stationary series on its past lags
and on past lags of error term.
• Look at ACF &PACF, Try a few likely
models
• Select one with lowest AIC, SIC
• Interpret R2, t and F as in CLRM
• Estimate the residuals necessary for
diagnostic testing
Diagnostic checking and
Forecasting
• Diagnostic check is to check for stationarity
of residuals
• Use ACF, PACF
• Box Pierce q and Ljung-Box LB statistics
• DF and ADF
• If no autocorrelation found then, the current
model may be used for forecasting.
• Forecasting requires undoing of the
“differences” taken to obtain stationarity
Forecasting
• To obtain forecast of level, rather than the first difference,
we need to integrate the first differenced series
• Eg: Forecasting model based on first differenced quarterly
GDP data upto 1991.4. Forecast for 1992.1

Y *
t  2 3 . 0 8  0 . 3 4 Y t * 1  0 . 2 9 Y t * 8  0 . 2 6 Y t * 1 2
Y 1 9 9 2. I  Y 1 9 9 1 . I V     1 ( Y 1 9 9 1 . I V  Y 1 9 9 1 . I I I )   8 ( Y 1 9 8 9 . I V  Y 1 9 8 9 . I I I )   1 2 ( Y 1 9 8 8. I V  Y 1 9 8 8 . I I I )  u 1 9 9 2. I
Y 1 9 9 2. I    ( 1   1 ) Y 1 9 9 1 . I V   1 Y 1 9 9 1 .I I I   8 Y 1 9 8 9 . I V   8 Y 1 9 8 9. I I I   1 2 Y 1 9 8 8 . I V   1 2 Y 1 9 8 8 . I I I  u 1 9 9 2. I

Y 1 9 9 2 .I  2 3 . 0 8  ( 1  0 . 3 4 ) Y 1 9 9 1 . I V  0 . 3 4 Y 1 9 9 1. I I I  0 . 2 9 Y 1 9 8 9. I V  0 . 2 9 Y 1 9 8 9. I I I  0 . 2 6 Y 1 9 8 8 . I V  0 . 2 6 Y 1 9 8 8 . I I I
• How will you integrate your model be to forecast for
1992.II?
• Using the same model above predict 1992.II
• (Data in page 794)
Vector Autoregression (VAR)
• VAR is similar to simultaneous equation model.
• It is a system of equations (vector)
• However VAR has no exogenous/ predetermined
variables
• If no predetermined variables, equation is
“unidentified” in simultaneous equation model
• VAR equation includes lagged values of all the
variables in the system
• Each equation is estimated independently and OLS
may be used
3 variable VAR model
m m m
Yt     iY t i   iX t i   iR t i   1t
i1 i 1 i1
m m m
X t   '    iY t i   i X t i    iR t i   2t
i 1 i1 i1
m m m
R t   ' '   iY t i    iX t i   iR t i  3t
i1 i1 i 1

 jt ' s c a lle d im p u ls e s o r in n o v a t io n s .
• Most critical part is to decide the no. of lags
– Too many lags leads to low df and multicollinearity
– Too few leads to specification errors
• AIC/SIC criterion may be used
• Since OLS is used interpretation same as CLRM
• However bcos of possibility of multicollinearity, t-test
should be less relied on. Rather consider F for joint
significance of the variables
VAR:Impulse response function & ECM
• IRF traces out the response of the dependent variable in
the VAR system to shocks in the error terms, such as u1,
u2 and u3.
• Logic being that suppose u1 in Y1 equation increases by a
value of 1 SD, such a shock will change Y1 in current and
future periods. Since Y1 enters the X and R equations,
change in u1 affects X and R too.
• IRF traces out this impact of such shocks for several
periods in the future.
• Impulse response functions are responses of all variables
in the model to a 1 unit structural shock to 1 variable in
the model.
• VECM estimates the short run impact just as in standard
ECM models.
VAR Limitations
• VAR model is a-theoretic.
• Less suited for policy analysis since emphasis is
forecasting.
• Problem of selecting lag length.
• In theory variables should be stationary, but many
use levels for interpretation.
• Quite often look at impulse response function
(IRF).
Bivariate Granger causality
• Determines direction of relationship of two variables (Ms and
GDP).
• Theoretically its very difficult to establish direction of
causation. Makes use of dictum that time does not run
backwards. Past lag can only affect present Present cannot
change past.
• Regress each variable with past lags of the itself and other
variable
• System of equation like simultaneous equations regressions.
• Unlike simultanoeus equation there are no exogenous
variables in the system
• More atheoretic compared to simultaneous eqn method
• Both variables are endogenous
• Both variables have to be tested first for cointegration
n n
GDPt    i M t i    j GDPt  j u1t
i 1 j 1
n n
M t   i M t i    j GDPt  j u 2t
i 1 j 1

• Unidirectional causality from M to GDP if if coefficients


on the lagged M are different zero as a group in eq 1 and
coefficient of GDP in eqn 2 is not different from zero
• Conversely unidirectional causality from GDP to M exists
if lagged GDP is significant in eqn 2 and coefficients of M
are not significant in eqn 1
• Bilateral or Feedback causality exists if coefficients of
lagged M and GDP are significant in both equations
• Independence is suggested when coefficients of lagged M
and GDP are not significant in both equations
Issues in estimation
• What is the lag size to be included
• AIC/SIC are used. Lags that give Minimum
values of AIC/SIC to be used
• It is atheoretic bcos no other control
variables can be included.
• Cointegration a must
• Only long run relationship indicated
• Inclusion of ECM to capture short run
dynamics not possible
• VAR a way out to address these issues
Measuring and Forecasting volatility
• Finance: Risk measurement
• Auto Regressive Conditional Heteroscedasticity ARCH:
Variance is not constant. Todays variance is influenced by
past variance plus shock term.
• GARCH: Generalised Auto Regressive Conditional
Heteroscedasticity
• Two ways of estimation of ARCH(p):
– Time series univariate approach: AR of variance
• ARCH(1) X 2
t    1X 2
 u
0 t 1 t

w h e r e Y t  d a i l y ` p r i c e `o f ` s t o c k ` A
Y t
*
 lo g Y t

dY t
*
 Yt*  Y *
t1
_
X t  dY t
*
 d Y t
*

– Fundamental theoretic approach: uses time series data


but k variable linear regression model
ARCH: k variable approach
Y t  1   2 X 2t  ..   k X kt ut
A R C H ( p )  u t2   0   1 u t2  1   2 u t 2 2  . .   k u 2
t p

• If coefficients are jointly significant, we have the ARCH


effect. There is clustering of volatility.

• GARCH(p,q):Variation of ARCH model.Conditional


variance of u at time t depends not only on the squared
error term in the previous time period (as in ARCH), but
also on conditional variance in previous period.
G A R C H ( 1 ,1 )     0   1 u
t
2 2
t 1   2 2
t1

w h e r e  t 2   0   1 u t2  1   2 u 2
t 2  . .   k u t2 p

G A R C H ( p , q )   t2   0   1 u t2 1  ..   p u t2  p   2  t 21  ..   q  t2 q

You might also like