Time Series in Matlab 14.

384 Time Series Analysis, Fall 2007 Recitation by Paul Schrimpf Supplementary to lectures given by Anna Mikusheva September 17, 2007

Recitation 2

Time Series in Matlab

clear; a = [1 0.5]; % AR coeffs b = [1 0.4 0.3]; % MA coeffs T = 1000; e = randn(T,1); % generate gaussian white noise y = filter(b,a,e); % generate y

The filter function can be used to generate data from an ARMA model, or apply a lter to a series.

Impulse-Response
To graph the impulse response of an ARMA, use fvtool
1 2

% create an impulse response fvtool(b,a,'impulse');

Sample Covariances
1 2 3 4

[c lags]=xcov(y,'biased'); figure; plot(lags,c); title('Sample Covariances');

The option, 'biased', says to use k = T k 1 k = T k t=1 (yt y )(ytk y )

y )(ytk y ); 'unbiased' would use

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Spectral Analysis

Spectral Analysis
Population spectral density for an ARMA:
1 2 3 4

% population density w = 0:0.1:pi; % frequencies to compute density at h = freqz(b,a,w); % returns frequency response = b(e{iw})/a(e{iw}) sd = abs(h).2./sqrt(2*pi); % make into density

Estimating the Spectral Density Parametric methods These estimate an AR(p) and use it to compute the spectrum.
1 2 3 4 5

[sdc wc] = pcov(y,8); % estimate spectral density by fitting AR(8) [sdy wy] = pyulear(y,8); % estimate spectral density by fitting AR(8) % using % the Yulewalker equations

Non-parametric methods
1 Denition 1. The sample periodogram of {xt }T is S() = T | t=1 T t=1

eit xt |2

Remark 2. The sample periodogram is equal to the Fourier transform of the sample autocovariances 1 S () = | T
T T 1

e
t=1

it

xt | =

k eik
k=T 1

[sdp wp] = periodogram(y,[],'onesided'); % estimate

using sample

Denition 3. A smoothed periodogram estimate of the spectral density is S () = 1 hT ( ) | T

t=1

eit xt |2 d

where hT () is some kernel weighting function. A smoothed periodogram is a weighting moving average of the sample periodogram. The following code estimates a smoothed periodogram using a Parzen kernel with band width T .

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Filtering

1 2

rT = round(sqrt(T)); [sdw ww] = pwelch(y,parzenwin(rT),rT1,[],'onesided'); % smoothed periodogram

Denition 4. A weighted covariance estimate of the spectrum is: S () = where gT (k) is some kernel.
ST

k gT (k)eik
k=ST

1 2 3 4 5 6 7 8 9 10

% bartlett weighted covariances wb = 0:0.1:pi; rT = sqrt(T); [c t]=xcov(y,'biased'); weight = 1abs(t)/rT; weight(abs(t)>rT) = 0; for j=1:length(wb) sdb(j) = sum(c.*weight.*exp(i*wb(j)*t)); end sdb = sdb / sqrt(2*pi);

Filtering
See: http://ideas.repec.org/c/wpa/wuwppr/0507001.html for code for common lters.

Example: Simulating an ARMA and estimating the spectrum

clear; close all; % closes all open figure windows % a b T e y model: y t = 0.9 y {t1} + b(L) e t = [1 0.7]; % AR coeffs = [1 0.3 2]; % MA coeffs = 200; = randn(T,1); % generate gaussian white noise = filter(b,a,e); % generate y

% plot y figure; plot(y); xlabel('t'); ylabel('y'); title('ARMA(1,2)');

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Example: Simulating an ARMA and estimating the spectrum

% create an impulse response fvtool(b,a,'impulse'); % calculate and plot sample autocovariances [c lags]=xcov(y,'biased'); figure; plot(lags,c); title('Sample Covariances'); % estimate spectral density % parametric [sdc wc] = pcov(y,8); % estimate spectral density by fitting AR(8) [sdy wy] = pyulear(y,8); % estimate spectral density by fitting AR(8) % using % the Yulewalker equations % nonparametric [sdp wp] = periodogram(y,[],'onesided'); % estimate using unsmoothed % periodogram rT = round(sqrt(T))*3; [sdw ww] = pwelch(y,parzenwin(rT),rT1,[],'onesided'); % smoothed periodogram % bartlett weighted covariances [c lags]=xcov(y,'biased'); t = (T1):(T1); weight = 1abs(t')/rT; weight(abs(t')>rT) = 0; wb = ww; for j=1:length(wb) sdb(j) = sum(c.*weight.*exp(i*wb(j)*((T1):(T1))')); end sdb = sdb / sqrt(2*pi); % population density w = wb; h = freqz(b,a,w); sd = abs(h).2./sqrt(2*pi); figure; plot(wp,sdp,wc,sdc,ww,sdw,wb,sdb,wy,sdy,w,sd); legend('raw periodogram','parametric AR','smoothed periodogram', ... 'bartlett weighted cov','YuleWalker','population density');

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Example: Simulating an ARMA and estimating the spectrum

ARMA(1,2) 15

10

10

15

50

100 t

150

200

Impulse Response

2.5

Amplitude

1.5

0.5

0 0 5 10 Samples 15 20 25

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Example: Simulating an ARMA and estimating the spectrum

Sample Covariances 16 14 12 10 8 6 4 2 0 2 4 200 150 100 50 0 50 100 150 200

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].

Example: Simulating an ARMA and estimating the spectrum

70 raw periodogram parametric AR smoothed periodogram bartlett weighted cov YuleWalker population density

60

50

40

30

20

10

0.5

1.5

2.5

3.5

Cite as: Paul Schrimpf, course materials for 14.384 Time Series Analysis, Fall 2007. MIT OpenCourseWare (http://ocw.mit.edu), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY].