12.07.2015 Views

"Frontmatter". In: Analysis of Financial Time Series

"Frontmatter". In: Analysis of Financial Time Series

"Frontmatter". In: Analysis of Financial Time Series

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

SIMPLE AUTOREGRESSIVE MODELS 39Model CheckingA fitted model must be examined carefully to check for possible model inadequacy.If the model is adequate, then the residual series should behave as a white noise.The ACF and the Ljung–Box statistics in Eq. (2.3) <strong>of</strong> the residuals can be used tocheck the closeness <strong>of</strong> â t to a white noise. For an AR(p) model, the Ljung–Boxstatistic Q(m) follows asymptotically a chi-squared distribution with m − p degrees<strong>of</strong> freedom. Here the number <strong>of</strong> degrees <strong>of</strong> freedom is modified to signify that pAR coefficients are estimated. If a fitted model is found to be inadequate, it must berefined.Consider the residual series <strong>of</strong> the fitted AR(3) model for the monthly valueweightedsimple returns. We have Q(10) = 15.8 with p value 0.027 based onits asymptotic chi-squared distribution with 7 degrees <strong>of</strong> freedom. Thus, the nullhypothesis <strong>of</strong> no residual serial correlation in the first 10 lags is rejected at the 5%level, but not at the 1% level. If the model is refined to an AR(5) model, then we haver t = 0.0092 + 0.107r t−1 − 0.001r t−2 − 0.123r t−3 + 0.028r t−4 + 0.069r t−5 +â t ,with ˆσ a = 0.054. The AR coefficients at lags 1, 3, and 5 are significant at the 5%level. The Ljung–Box statistics give Q(10) = 11.2 with p value 0.048. This modelshows some improvements and appears to be marginally adequate at the 5% significancelevel. The mean <strong>of</strong> r t based on the refined model is also very close to 0.01,showing that the two models have similar long-term implications.2.4.3 ForecastingForecasting is an important application <strong>of</strong> time series analysis. For the AR(p) modelin Eq. (2.7), suppose that we are at the time index h and are interested in forecastingr h+l ,wherel ≥ 1. The time index h is called the forecast origin and the positiveinteger l is the forecast horizon.Letˆr h (l) be the forecast <strong>of</strong> r h+l using the minimumsquared error loss function. <strong>In</strong> other words, the forecast ˆr k (l) is chosen such thatE[r h+l −ˆr h (l)] 2 ≤ min gE(r h+l − g) 2 ,where g is a function <strong>of</strong> the information available at time h (inclusive). We referredto ˆr h (l) as the l-step ahead forecast <strong>of</strong> r t at the forecast origin h.1-Step Ahead ForecastFrom the AR(p) model, we haver h+1 = φ 0 + φ 1 r h +···+φ p r h+1−p + a h+1 .Under the minimum squared error loss function, the point forecast <strong>of</strong> r h+1 given themodel and observations up to time h is the conditional expectation

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!