Download presentation
Presentation is loading. Please wait.
Published byAldous Campbell Modified over 8 years ago
1
Analysis of Financial Data Spring 2012 Lecture 5: Time Series Models - 3 Priyantha Wijayatunga Department of Statistics, Umeå University Priyantha.wijayatunga@stat.umu.se Course homepage: http://www8.stat.umu.se/kursweb/vt012/staa2st017mom2/ http://www8.stat.umu.se/kursweb/vt012/staa2st017mom2/
2
Some Demonstrations
3
ACF and PACF ACF and PACF (that cuts-off at lag 1 ) look like if the AR(1) model can fit the data
4
Best Fitting Model Stationary R-squared should be big! Check the significance of the residual autocorrelation with the Ljung–Box test Model Statistics a ModelNumber of Predictors Model Fit statisticsLjung-Box Q(18) Number of Outliers Stationary R-squaredStatisticsDFSig. x-Model_10,61420,82217,2340 a. Best-Fitting Models according to Stationary R-squared (larger values indicate better fit). ARIMA Model Parameters a EstimateSEtSig. x-Model_1xNo TransformationConstant50,239,203247,646,000 ARLag 1,783,01456,311,000 a. Best-Fitting Models according to Stationary R-squared (larger values indicate better fit). Model Description Model Type Model IDxModel_1ARIMA(1,0,0)
5
Another Time Series
6
ACF and PACF ACF and PACF (that cuts-off at lag 2 ) show that if AR(2) model can fit the data
7
Best Fitting Model Stationary R-squared should be big! Check the significance of the residual autocorrelation with the Ljung–Box test Model Description Model Type Model IDxModel_1ARIMA(2,0,0) Model Statistics Model Number of Predictors Model Fit statisticsLjung-Box Q(18) Number of Outliers Stationary R- squaredRMSEStatisticsDFSig. x-Model_11,8181,94919,14716,2610 ARIMA Model Parameters EstimateSEtSig. x-Model_1xNo TransformationConstant48,829,99449,121,000 ARLag 1,810,02236,359,000 Lag 2,104,0224,685,000 DAY, not periodicNo TransformationNumeratorLag 0,001 1,319,187
8
Residual ACF and PACF
9
Another Time Series
10
ACF of Time Series Since ACF is positive until large lags, it is an indication of nonstationarity. Differencing is needed
11
ACF and PACF of 1-Differenced Time Series
12
Model Description Model Type Model IDxModel_1ARIMA(2,1,0) Model Statistics ModelNumber of Predictors Model Fit statisticsLjung-Box Q(18) Number of Outliers Stationary R-squaredRMSEStatisticsDFSig. x-Model_10,7791,98011,83516,7550 ARIMA Model Parameters EstimateSEtSig. x-Model_1xNo TransformationConstant,248,418,593,553 ARLag 1,781,02235,119,000 Lag 2,114,0225,110,000 Difference1
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.