A First Course on Time Series Analysis : Examples with SAS by Michael Falk

By Michael Falk

Similar analysis books

Econometric Analysis of Financial Markets

This selection of papers represents the cutting-edge within the applicationof contemporary econometric easy methods to the research of economic markets. From a methodological standpoint the most emphasis is on cointegration research and ARCH modelling. In cointegration research the hyperlinks among long-runcomponents of time sequence are studied.

Magnetic Domains: The Analysis of Magnetic Microstructures

The wealthy international of magnetic microstructure or magnetic domain names - extending from the "nano-world" to noticeable dimensions - is systematically coated during this publication. the topic, that may be known as "mesomagnetism", kinds the hyperlink among atomic foundations and technical purposes of magnetic fabrics, starting from laptop garage platforms to the magnetic cores of electric equipment.

Additional resources for A First Course on Time Series Analysis : Examples with SAS

Example text

Z − zp ) z z z =c 1− 1− ... 1 − , z1 z2 zp where c := ap (−1)p z1 . . zp . In case of |zi | > 1 we can write for |z| < |zi | 1 1 u u = z , 1 − zzi z i u≥0 where the coefficients (1/zi )u , u ≥ 0, are absolutely summable. In case of |zi | < 1, we have for |z| > |zi | 1 1− 1 z zi =−z zi 1 1− zi z zi =− z ziu z −u u≥0 =− u≤−1 1 zi u zu, where the filter with coefficients −(1/zi )u , u ≤ −1, is not a causal one. In case of |zi | = 1, we have for |z| < 1 1 1− z zi = u≥0 1 zi u zu, where the coefficients (1/zi )u , u ≥ 0, are not absolutely summable.

Difference Filter We have already seen that we can remove a periodic seasonal component from a time series by utilizing an appropriate linear filter. We 28 Elements of Exploratory Time Series Analysis will next show that also a polynomial trend function can be removed by a suitable linear filter. 5. For a polynomial f (t) := c0 + c1 t + · · · + cp tp of degree p, the difference ∆f (t) := f (t) − f (t − 1) is a polynomial of degree at most p − 1. Proof. The assertion is an immediate consequence of the binomial expansion p p (t − 1) = k=0 p k t (−1)p−k = tp − ptp−1 + · · · + (−1)p .

10. Let a ∈ C. The filter (au ) with a0 = 1, a1 = −a and au = 0 elsewhere has an absolutely summable and causal inverse filter (bu )u≥0 if and only if |a| < 1. In this case we have bu = au , u ≥ 0. Proof. The characteristic polynomial of (au ) is A1 (z) = 1−az, z ∈ C. Since the characteristic polynomial A2 (z) of an inverse filter satisfies A1 (z)A2 (z) = 1 on some annulus, we have A2 (z) = 1/(1−az). Observe now that 1 = au z u , if |z| < 1/|a|. , u≥0 |au | < ∞. If |a| ≥ 1, then A2 (z) = u≥0 au z u exists for all |z| < 1/|a|, but u≥0 |a|u = ∞, which completes the proof.