Scribd
Upload
8 views11 pages

Predicting Asset Prices

This document discusses different time series models for predicting asset prices: 1) Time series can be stationary or non-stationary, with the latter containing trends or seasonality that require differencing to make stationary. 2) Autoregressive (AR) models use past values to predict future values assuming stationarity. 3) Moving average (MA) models use past forecast errors rather than past values. 4) Autoregressive integrated moving average (ARIMA) models combine AR, differencing, and MA components. 5) Seasonal ARIMA (SARIMA) models extend ARIMA to account for seasonal patterns with an additional seasonal backshift parameter.

Uploaded by

akpanyap
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
8 views11 pages

Predicting Asset Prices

This document discusses different time series models for predicting asset prices: 1) Time series can be stationary or non-stationary, with the latter containing trends or seasonality that require differencing to make stationary. 2) Autoregressive (AR) models use past values to predict future values assuming stationarity. 3) Moving average (MA) models use past forecast errors rather than past values. 4) Autoregressive integrated moving average (ARIMA) models combine AR, differencing, and MA components. 5) Seasonal ARIMA (SARIMA) models extend ARIMA to account for seasonal patterns with an additional seasonal backshift parameter.

Uploaded by

akpanyap
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/ 11

PREDICTING ASSET PRICES

PRICES/RETURNS
1. Time series
• Stationary
• No identifiable long-run patterns
• Variation of values are random
• Non-stationary: may contain two types of patterns
• Trend:- long-term increase or decrease of the series
• Seasonality:- time-based pattern of fluctuation in the series
• Patterns are always of a constant period, e.g. quarterly, monthly or weekly
• However if the fluctuation is not constant, then the pattern is called cyclicality.
cont.
• Transforming non-stationary time series to stationary time series
• Use the Differencing method
• This method accounts for the non-stationary by subtracting the value of a
previous observation from the value of consecutive observation of the series
• This will be done in two or more differencing depending on the data
• First-Order differencing
• Second-Order differencing
First-Order differencing

𝑦′ 𝑡 =𝑦 𝑡 −𝑦 𝑡 −1
Original Series
2
4
6
8
10
12
First-Order Differenced Series

(4-2)=2
(6-4)=2
(8-6)=2
(10-8)=2
(12-10)=2
Second-Order Differencing
• Sometimes, the data will need further differencing to be made
stationary

Original Series First-Order Differenced Series Second-Order Differenced Series


2
4 (4-2)=2
8 (8-4)=4 (4-2)=2
14 (14-8)=6 (6-4)=2
22 (22-14)=8 (8-6)=2
32 (32-22)=10 (10-8)=2
Seasonal Differencing
• Where fluctuations still exist after the second differencing, the
seasonal differencing is conducted.

• Where m is the number of lags corresponding to the season.


• Example if we were looking for daily fluctuations in an hourly time
series, m=24
• Thus, First-Order, Second-Order and Seasonal differencing can be
used independently or together.
2. Autoregressive Model (AR)
• Follows the assumption that past observations of the time series can be
used to predict the current or future value.
• Assumptions
• Data are stationary
• It is a regression of variables against past observations of itself.
• The model is as follows:

Where is the current value, c is a constant, , , … are previous


observations of the value at times , … and
is the white noise (i.e. a random error).
3. Moving Average Model (MA)
• Instead of using the past values of the variable, we use past errors( of
the forecast.
• Any given value () can be quantified by a mean value and weighted
moving average of past forecast errors as follows:

• Where is the current value, c is the mean value of the series and , , …
are the forecast errors at times t, , …
4. Autoregressive (Integrated) Moving
Average Model (ARMA/ARIMA)
• An autoregressive (integrated) moving average model includes
autoregressive, moving average and differencing.
• The ARIMA (p,d,q) model is as follows:

• Where is the differenced series, is the AR component and is the MA


component of the lagged series.
5. Seasonal Autoregressive Integrated Moving Average
Model
• An ARIMA model is unable to take into account seasonal differencing
because d accounts for only orders of differencing.
• A seasonal autoregressive integrated moving average (SARIMA)
model is similar to ARIMA, with an additional backshift of the
seasonal period.
• The notation for this is as follows:

• With being the non-seasonal part, and the number of periods per
season.
THANK YOU
QUESTIONS???

You might also like