Arima 1

AUTOREGRESSIVE
MOVINGAVERAGE (ARMA)
PROF. AJAYA KUMAR PANDA

Test of Stationarity/Unit root:
➢ A time series where mean, variance and covariance are time invariant is said to be
stationary (or) covariance stationary (or) weakly stationary.
➢ The data which do not fulfill these properties are called non-stationary. A non-stationary
process is also called as a unit root process.
➢ The econometric models using non-stationary data are likely to violate the desirable
statistical properties of the estimators or may give misleading inferences. Thus, it is
necessary to test the stationarity of the time series before attempting any econometric
exercise.
How to Test Stationary:
There are both formal and informal tests of stationarity.

The informal tests include time series plots and use of correlogram. Statistical packages
use Box-Pierce Q-statistics and L-jung-Box Q-statistics for testing stationarity of series.
These two statistics are based on autocorrelation coefficient of several lag lengths.
The formal tests of non-stationarity are also known as test of unit root. Dickey-Fuller
(DF) and Augmented Dickey-Fuller (ADF) test are popular methods to check the
presence of unit root in the data.
TESTING STATIONARY USING
CORRELOGRAM
H0 : Variable is stationary or there is no trend

H1 : Variable is not a stationary or there is trend
 If probability of Q- stat of ACF and PACF is less than 0.05 then reject H0.
 In case of correlogram, we wish not to reject H0 and expect the probability values to
be greater than 0.05
 ACF: Auto-Correlation Function (ACF) represents the correlation between the
observation at the current time ‘t’ and it's lag (t – i). For e.g., if we assume that up-to day
‘i’ stock prices are correlated with its past values then we can calculate its ACF to know
how effectively today's stock price is correlated with it’s past.
 PACF: Partial Auto-Correlation function (PACF) represents the correlation between
observation at two time period given that we consider both the observations are correlated
to the observations at other time period. In other words, PACF measures the correlation
between time series observations after controlling for the correlations at intermediate lags.
Hence PACF measures the true marginal effects of significant lags.
➢ Use ACF to determine the terms of MA. We must find significant order of ACF to
identify optimum order of MA(q).
➢ Use PACF to determine the terms of AR. Like previous, We must find significant order
of ACF to identify optimum order of AR(p).
➢ Similarly, looking at PACF and ACF together, we will find the order of ARMA (p, q).
Note: The dotted lines present the significant thresholds. The bars/lines represents ACF and PACF values
at each time lag. Only the Bars/lines that cross the significant threshold lines or confidence interval are
called significant.
H0: Variable is stationary or there is no trend.
➢ This is the common pattern of non-stationary time series

➢ ACF lies outside the CI lines and declines slowly. It shows high autocorrelation in the errors.
➢ PACF drops immediately after 1st lag, but most of the PACFs are having prob. < 0.05 even lying
within the CI (i.e. dotted lines).
Pattern of Stationary Data
Note:
➢ This is the common pattern of a stationary series
➢ ACF and PACF are within the CI and the prob. values are greater than 0.05. Hence, we don’t
reject the H0.
Let’s Summarize the pattern of ACF and PACF in ARMA modelling.

ACF Pattern PACF Pattern
AR (P) → Exponential decay (or) Significant spikes through 1st lag
Damped sine wave pattern
MA (q) → Significant spikes through 1st Exponential decay (or)

lag Damped sine wave pattern
ARMA (1,1) → Exponential Decay from Exponential Decay from lag 1
lag 1
ARMA (p, q) → Exponential Decay Exponential Decay
Formal Method: Dickey- Fuller (DF) Test:
Assumption:
▪ DF confines itself to pure Autoregressive Integrated Moving Average
process of order (ARIMA: 1, 0, 0)
▪ DF/ADF test assumes that the error process is independently and identically
distributed (0, σ2) i.e. a Gaussian processes.
DF/ADF consider three different regression equations
Yt= βYt-1 + εt A pure random walk process
Yt= α0 + βYt-1 + εt A pure random walk with intercept
Yt= α0 + α1 t + βYt-1 + εt A pure random walk with intercept and linear time trend
Steps to model AR, MA, ARMA:
Step 1: Identification process. That is analyze the time series plot to
visualize stationarity, trend, Seasonality etc.
Step 2: Undertake testing of unit root test through DF/ADF/PP/KPSS etc.
Step 3: Analyze both at ACF & PACF for the data both at level and 1st the
data difference
Step 4: Decide order of AR and MA and finalize the possible of ARMA
(p, q) models.
Step 5: Estimate the models
AR, MA, ARMA
(1) 𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝛽2 𝑌𝑡−2 + 𝜀𝑡 AR(2)
(2) 𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝛽2 𝑌𝑡−2 + 𝛽1 𝑌𝑡−3 + 𝛽2 𝑌𝑡−4 + 𝜀𝑡 AR(4)
(3) 𝜀𝑡 = 𝛼 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝜇𝑡 MA(2)
(4) 𝜀𝑡 = 𝛼 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝛾3 𝜀𝑡−3 + 𝛾4 𝜀𝑡−4 + 𝜇𝑡 MA(4)
(5) 𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝛽2 𝑌𝑡−2 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝛾3 𝜀𝑡−3 + 𝛾4 𝜀𝑡−4 +

𝜀𝑡
ARMA(2,4)
𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝜀𝑡 ……… (1)
𝑌𝑡−1 = 𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ……….. (2)
𝑌𝑡 = 𝛼 + 𝛽1 (𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ) + 𝜀𝑡 ……… (3)
𝑌𝑡 = (𝛼 + 𝛽1 𝐴) + (𝛽1 𝐵1 𝑌𝑡−2 ) + [(𝛽1 𝐸𝑡 ) + 𝜀𝑡 ]
𝑌𝑡 = 𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ……… (4)

Identifying the Best Model:
Select the model on the basis of below characteristics
➢ Maximum number of significant coefficients

➢ The model with least volatility (Sigma)
➢ The model with lowest AIC or SBC value
➢ The model with highest Adj. R2
Probable Models
ARMA ARMA ARMA ARMA ARMA
AR(2) AR(3) (1,0,1) (2,0,1) (3,0,1) (1,0,2) (2,0,3)
Sig. Coeff 2 1 2 3 2 2 3
Sigma 0.69 0.83 0.34 0.38 0.53 0.64 0.91
Sigma2 0.47 0.69 0.11 0.15 0.28 0.40 0.82
R2 0.80 0.83 0.90 0.94 0.91 0.89 0.90
Adj. R2 0.78 0.81 0.88 0.92 0.89 0.84 0.88
AIC 1323.14 1113.14 1033.14 1003.11 1123.14 1163.14 1128.87
SBIC 1142.49 1145.65 1152.94 1129.14 1121.44 1132.44 1141.09
Diagnostic test of the best ARMA Model
To cross verify the best ARMA model, that you have selected on the basis of the
above four selection criteria, follow the following steps:
➢ Estimate the selected model.
➢ Estimate or predict the estimated error distribution on the ARMA model
➢ Then calculate the correlogram of the estimated errors.
➢ If the ACF and PACF of the estimated residual is within 95% CI, dotted line
and prob. values are greater than 0.05, we conclude that ARMA model is best .
➢ If any of the ACF or PACF are crossing the lines and prob < 0.05 then we may
conclude that the selected model could not capture some relevant information
that are left with random errors.
Thank You

Arima 1

Uploaded by

Copyright:

Available Formats

Arima 1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Arima 1

Uploaded by

Copyright:

Available Formats

AUTOREGRESSIVE

PROF. AJAYA KUMAR PANDA

There are both formal and informal tests of stationarity.

H0 : Variable is stationary or there is no trend

H0: Variable is stationary or there is no trend.

➢ This is the common pattern of non-stationary time series

Let’s Summarize the pattern of ACF and PACF in ARMA modelling.

MA (q) → Significant spikes through 1st Exponential decay (or)

Yt= α0 + βYt-1 + εt A pure random walk with intercept

Step 2: Undertake testing of unit root test through DF/ADF/PP/KPSS etc.

(2) 𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝛽2 𝑌𝑡−2 + 𝛽1 𝑌𝑡−3 + 𝛽2 𝑌𝑡−4 + 𝜀𝑡 AR(4)

(3) 𝜀𝑡 = 𝛼 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝜇𝑡 MA(2)

(4) 𝜀𝑡 = 𝛼 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝛾3 𝜀𝑡−3 + 𝛾4 𝜀𝑡−4 + 𝜇𝑡 MA(4)

(5) 𝑌𝑡 = 𝛼 + 𝛽1 𝑌𝑡−1 + 𝛽2 𝑌𝑡−2 + 𝛾1 𝜀𝑡−1 + 𝛾2 𝜀𝑡−2 + 𝛾3 𝜀𝑡−3 + 𝛾4 𝜀𝑡−4 +

𝑌𝑡−1 = 𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ……….. (2)

𝑌𝑡 = 𝛼 + 𝛽1 (𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ) + 𝜀𝑡 ……… (3)

𝑌𝑡 = (𝛼 + 𝛽1 𝐴) + (𝛽1 𝐵1 𝑌𝑡−2 ) + [(𝛽1 𝐸𝑡 ) + 𝜀𝑡 ]

𝑌𝑡 = 𝐴 + 𝐵1 𝑌𝑡−2 + 𝐸𝑡 ……… (4)

➢ Maximum number of significant coefficients

You might also like