Scribd
Upload
84 views12 pages

Some Examples of Non-Stationary Time Series:: Model ACF Pacf

This document provides examples of different time series models and describes their expected autocorrelation (ACF) and partial autocorrelation (PACF) functions. It includes examples of white noise, random walks, linear trends, seasonal cycles, autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) processes. For each model, it gives the mathematical equation and examines the corresponding ACF and PACF graphs to help identify the characteristics of different time series models.

Uploaded by

Tanmay Maity
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
84 views12 pages

Some Examples of Non-Stationary Time Series:: Model ACF Pacf

This document provides examples of different time series models and describes their expected autocorrelation (ACF) and partial autocorrelation (PACF) functions. It includes examples of white noise, random walks, linear trends, seasonal cycles, autoregressive (AR), moving average (MA), and autoregressive moving average (ARMA) processes. For each model, it gives the mathematical equation and examines the corresponding ACF and PACF graphs to help identify the characteristics of different time series models.

Uploaded by

Tanmay Maity
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

Some examples of Non-stationary time series:

Behavior of ACF and PACF for ARMA model:


Model ACF PACF

White Noise All zeros All zeros

AR(p) Exponential Decay p significant lags before dropping to zero

MA(q) q significant lags before dropping to zero Exponential Decay

ARMA(p,q) Decay after qth lag Decay after pth lag


White Noise process:

𝑦𝑡 ~ 𝑁(0, 5)

ACF and PACF Function of White Noise process: 𝑦𝑡 ~ 𝑁(0, 5)


Random Walk model without drift:

𝑦𝑡 = 𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = 10

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝑅𝑎𝑛𝑑𝑜𝑚 𝑊𝑎𝑙𝑘 𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑑𝑟𝑖𝑓𝑡 [𝑦𝑡 = 𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = 10]
Random Walk model with drift:
(Stochastic trend)

𝑦𝑡 = 𝑎 + 𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = 10, a = .9

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝑅𝑎𝑛𝑑𝑜𝑚 𝑊𝑎𝑙𝑘 𝑤𝑖𝑡ℎ 𝑑𝑟𝑖𝑓𝑡 [𝑦𝑡 = 𝑎 + 𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = 10, a = .9

]
Model with Liner trend:

𝑦𝑡 = .5 + .6𝑡 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝐿𝑖𝑛𝑒𝑎𝑟 𝑡𝑟𝑒𝑛𝑑 𝑚𝑜𝑑𝑒𝑙 [𝑦𝑡 = .5 + .6𝑡 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]
Model with Linear combination of harmonic terms/ cyclical component

2𝜋𝑡 2𝜋𝑡
[𝑦𝑡 = 𝑎 + 𝑏𝐶𝑜𝑠 ( ) + 𝑐𝑆𝑖𝑛( ); a=5, b=.5, c=.5, 𝜆 = 80]
𝜆 𝜆

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 the Model with Linear combination of cyclical components
2𝜋𝑡 2𝜋𝑡
[𝑦𝑡 = 𝑎 + 𝑏𝐶𝑜𝑠 ( 𝜆 ) + 𝑐𝑆𝑖𝑛( 𝜆 ); a=5, b=.5, c=.5, 𝜆 = 80]
Autoregressive model
AR(1) Model: Positive coefficient

𝑦𝑡 =. 8𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 5), 𝑦0 = 15]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝐴𝑅 (1)𝑚𝑜𝑑𝑒𝑙 𝑤𝑖𝑡ℎ 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 [𝑦𝑡 =. 8𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 5), 𝑦0 = 15]
]
AR(1) Model: Negative coefficient

𝑦𝑡 = −. 8𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 5), 𝑦0 = 15]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝐴𝑅 (1)𝑚𝑜𝑑𝑒𝑙 𝑤𝑖𝑡ℎ 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 [𝑦𝑡 = −. 8𝑦𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 5), 𝑦0 = 15]

]
AR (2):

𝑦𝑡 =. 8𝑦𝑡−1 −. 6𝑦𝑡−2 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = .1, 𝑦1 = .1]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝐴𝑅 (2)𝑚𝑜𝑑𝑒𝑙 [𝑦𝑡 =. 8𝑦𝑡−1 −. 6𝑦𝑡−2 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2), 𝑦0 = .1, 𝑦1 = .1]

]
MA(1) Model: Positive coefficient

𝑦𝑡 =. 75𝜀𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝑀𝐴(1)𝑚𝑜𝑑𝑒𝑙 𝑤𝑖𝑡ℎ 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 [𝑦𝑡 =. 75𝜀𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

]
MA(1) model: Negative coefficient

𝑦𝑡 = −. 75𝜀𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝑀𝐴(1)𝑚𝑜𝑑𝑒𝑙 𝑤𝑖𝑡ℎ 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 [𝑦𝑡 = −. 75𝜀𝑡−1 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

]
]
MA(2) model:

𝑦𝑡 =. 75𝜀𝑡−1 + .9𝜀𝑡−2 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

𝐴𝐶𝐹 𝑎𝑛𝑑 𝑃𝐴𝐶𝐹 𝑜𝑓 𝑀𝐴(2)𝑚𝑜𝑑𝑒𝑙 [𝑦𝑡 =. 75𝜀𝑡−1 + .9𝜀𝑡−2 + 𝜀𝑡 , 𝜀𝑡 ~𝑁(0, 2)]

You might also like