Linear Filters
denote a bivariate time series with zero mean. Let
The time series {y t : t T} is said to be constructed from {x t : t T} by means of a Linear Filter. Suppose that the time series {y t : t T} is constructed as follows:
The autocovariance function of the filtered series
Thus the spectral density of the time series {y t : t T} is:
Comment A: is called the Transfer function of the linear filter. is called the Gain of the filter while is called the Phase Shift of the filter.
Also
Thus cross spectrum of the bivariate time series is:
Definition: = Squared Coherency function Note:
Comment B: = Squared Coherency function. if {y t : t T} is constructed from {x t : t T} by means of a linear filter
Linear Filters with additive noise at the output
denote a bivariate time series with zero mean. Let t =..., -2, -1, 0, 1, 2,... Suppose that the time series {y t : t T} is constructed as follows: The noise {v t : t T} is independent of the series {x t : t T} (may be white)
t
The autocovariance function of the filtered series with added noise
continuing Thus the spectral density of the time series {y t : t T} is:
Also
Thus cross spectrum of the bivariate time series is:
Thus = Squared Coherency function. Noise to Signal Ratio
Box-Jenkins Parametric Modelling of a Linear Filter
Consider the Linear Filter of the time series {X t : t T}: where and = the Transfer function of the filter.
{a t : t T} is called the impulse response function of the filter since if X 0 =1and X t = 0 for t ≠ 0, then : for t T Linear Filter XtXt atat
Also Note:
Hence { Y t } and { X t } are related by the same Linear Filter. Definition The Linear Filter is said to be stable if : converges for all |B| ≤1.
Discrete Dynamic Models:
Many physical systems whose output is represented by Y(t) are modeled by the following differential equation: Where X(t) is the forcing function.
If X and Y are measured at discrete times this equation can be replaced by: where = I-B denotes the differencing operator.
This equation can in turn be represented with the operator B. orwhere
This equation can also be written in the form as a Linear filter as Stability: It can easily be shown that this filter is stable if the roots of (x) = 0 lie outside the unit circle.
Determining the Impulse Response function from the Parameters of the Filter:
Now or Hence
Equating coefficients results in the following conclusions: a j = 0 for j < b. a j - 1 a j-1 - 2 a j r a j-r = j ora j = 1 j-1 + 2 a j r a j-r + j for b ≤ j ≤ b+s. anda j - 1 a j-1 - 2 a j r a j-r = 0 ora j = 1 a j-1 + 2 a j r a j-r for j > b+s.
Thus the coefficients of the transfer function, a 0, a 1, a 2,..., satisfy the following properties 1)b zeroesa 0, a 1, a 2,..., a b-1 2)No pattern for the next s-r+1 values a b, a b+1, a b+2,..., a b+s-r 3)The remaining values a b+s-r+1, a b+s-r+2, a b+s-r+3,... follow the pattern of an r th order difference equation a j = 1 a j-1 + 2 a j r a j-r
Exampler =1, s=2, b=3, 1 = a 0 = a 1 = a 2 = 0 a 3 = a 2 + 0 = 0 a 4 = a 3 + 1 = 0 + 1 a 5 = a 4 + 2 = [ 0 + 1 ] + 2 = 2 w 0 + 1 + 2 a j = a j-1 for j ≥ 6.
Transfer function {a t }
Identification of the Box-Jenkins Transfer Model with r=2
Recall the solution to the second order difference equation a j = 1 a j-1 + 2 a j-2 follows the following patterns: 1)Mixture of exponentials if the roots of 1 - 1 x - 2 x 2 = 0 are real. 2) Damped Cosine wave if the roots to 1 - 1 x - 2 x 2 = 0 are complex. These are the patterns of the Impulse Response function one looks for when identifying b,r and s.
Estimation of the Impulse Response function, a j (without pre-whitening).
Suppose that {Y t : t T} and {X t : t T}are weakly stationary time series satisfying the following equation: Also assume that {N t : t T} is a weakly stationary "noise" time series, uncorrelated with {X t : t T}. Then
Suppose that for s > M, a s = 0. Then a 0, a 1,...,a M can be found solving the following equations:
If the Cross autocovariance function, XY (h), and the Autocovariance function, XX (h), are unknown they can be replaced by their sample estimates C XY (h) and C XX (h), yeilding estimates of the impluse response function
In matrix notation this set of linear equations can be written:
If the Cross autocovariance function, XY (h), and the Autocovariance function, XX (h), are unknown they can be replaced by their sample estimates C XY (h) and C XX (h), yeilding estimates of the impluse response function
Estimation of the Impulse Response function, a j (with pre-whitening).
Suppose that {Y t : t T} and {X t : t T}are weakly stationary time series satisfying the following equation: Also assume that {N t : t T} is a weakly stationary "noise" time series, uncorrelated with {X t : t T}.
In addition assume that {X t : t T}, the weakly stationary time series has been identified as an ARMA(p,q) series, estimated and found to satisfy the following equation: (B)X t = (B)u t where {u t : t T} is a white noise time series. Then [ (B)] -1 (B)X t = u t transforms the Time series {X t : t T} into the white noise time series{u t : t T}.
This process is called Pre-whitening the Input series. Applying this transformation to the Output series {Y t : t T} yeilds:
or where and
In this case the equations for the impulse response function - a 0, a 1,...,a M - become (assuming that for s > M, a s = 0):
Summary Identification and Estimation of Box-Jenkins transfer model
To identify the series we need to determine b, r and s. The first step is to compute 1.the ACF’s and the cross CF’s C xx (h) and C xy (h) 2.Estimate the impulse response function using
The Impulse response function {a t } bs- r + 1 Pattern of an r th order difference equation 3.Determine the value of b, r and s from the pattern of the impulse response function
3.Determine preliminary estimates of the Box- Jenkins transfer function parameters using: i.for j > b+s.. a j = 1 a j-1 + 2 a j r a j-r ii.for b ≤ j ≤ b+s a j = 1 j-1 + 2 a j r a j-r + j 4.Determine preliminary estimates of the ARMA parameters of the input time series {x t }
5.Determine preliminary estimates of the ARIMA parameters of the noise time series { t }
Maximum Likelihood estimation of the parameters of the Box-Jenkins Transfer function model
The Box- Jenkins model is written The parameters of the model are: In addition 1.the ARMA parameters of the input series {x t } 2.The ARIMA parameters of the noise series { t }
The model for the noise { t }series can be written
Given starting values for {y t }, {x t }, and and the parameters of the transfer function model and the noise model We can calculate successively: The maximum likelihood estimates are the values that minimize:
Fitting a transfer function model Example: Monthly Sales (Y) and Monthly Advertising expenditures
The Data
Using SAS Available in the Arts computer lab
The Start up window for SAS
To import data - Choose File -> Import data
The following window appears
Browse for the file to be imported
Identify the file in SAS
The next screen (not important) click Finish
The finishing screen
You can now run analysis by typing code into the Edit window or selecting the analysis form the menu To fit a transfer function model we need to identify the model –Determine the order of differencing to achieve Stationarity –Determine the value of b, r and s.
To determine the degree of differencing we look at ACF’s and PACF’s for various order of differencing
To produce the ACF, PACF – type the following commands into the Editor window- Press Run button
To identify the transfer function model we need to estimate the impulse response function using: For this we need the ACF of the input series and the cross ACF of the input with the output
To produce the Cross correlation function – type the following commands into the Editor window
the impulse response function using can be determined using some other package (i.e. Excel) b = 4 r,s = 1
To Estimate the transfer function model – type the following commands into the Editor window
To estimate the following model Use input=( b $ ( -lags ) / ( -lags) x) In SAS
The Output