|
==Description==
Given a time series <math>z_t</math>, the partial autocorrelation of lag ''<math>k''</math>, denoted <math>\alpha(k)phi_{kk}</math>, is the [[autocorrelation]] between <math>z_t</math> and <math>z_{t+k}</math> with the linear dependence of <math>z_t</math> on <math>z_{t+1}</math> through <math>z_{t+k-1}</math> removed; equivalently, it is the autocorrelation between <math>z_t</math> and <math>z_{t+k}</math> that is not accounted for by lags <math>1</math> through <math>k-1</math>, inclusive.
: <math display="block">\alpha(1)phi_{11} = \operatorname{corr}(z_{t+1}, z_{t}),\text{ for }k= 1,</math><math display="block">\phi_{kk} = \operatorname{corr}(z_{t+k} - \hat{z}_{t+k},\, z_{t} - \hat{z}_{t}),\text{ for }k\geq 2.</math>
: <math>\alpha(k) = \operatorname{corr}(z_{t+k} - \hat{z}_{t+k},\, z_{t} - \hat{z}_{t}),\text{ for }k\geq 2.</math>
where <math>\hat{z}_{t+k} = \beta_1 z_{t+k-1} + \beta_2 z_{t+k-2} + ... + \beta_{k-1} z_{t+1}</math> is the [[linear combination]] of <math>\{z_{t+k-1}, z_{t+k-2}, ..., z_{t+1}\}</math> that minimizes the [[mean squared error]], <math>\Epsilon[z_{t+k} - \hat{z}_{t+k}]^2</math>. Similarly, <math>\hat{z}_t = \beta_1 z_{t+1} + \beta_2 z_{t+2} + ... + \beta_{k-1} z_{t+k-1} </math> is a linear combination minimizing <math>\Epsilon[z_t - \hat{z}_t]^2</math>. For [[Stationary process|stationary processes]], the coefficients <math>\beta_1, \beta_2, ..., \beta_{k-1} </math> are the same.<ref>{{Cite book |last=Shumway |first=Robert H. |url=http://link.springer.com/10.1007/978-3-319-52452-8 |title=Time Series Analysis and Its Applications: With R Examples |last2=Stoffer |first2=David S. |date=2017 |publisher=Springer International Publishing |isbn=978-3-319-52451-1 |series=Springer Texts in Statistics |___location=Cham |pages=97-98 |language=en |doi=10.1007/978-3-319-52452-8}}</ref>
There are algorithms for estimating the partial autocorrelation based on the sample autocorrelations.<ref name=":0">{{Cite book |last=Box |first=George E. P. |title=Time Series Analysis: Forecasting and Control |last2=Reinsel |first2=Gregory C. |last3=Jenkins |first3=Gwilym M. |publisher=John Wiley |year=2008 |isbn=9780470272848 |edition=4th |___location=Hoboken, New Jersey |language=en}}</ref><ref>{{Cite book |last=Brockwell |first=Peter J. |title=Time Series: Theory and Methods |last2=Davis |first2=Richard A. |publisher=Springer |year=1991 |isbn=9781441903198 |edition=2nd |___location=New York, NY |language=en}}</ref> TheseOne algorithmsof derivethese fromprocedures is the exact[[Levinson-Durbin|Levinson–Durbin theoreticalAlgorithm]]. relationThe betweenpartial autocorrelation of any time series can be calculated by iteratively solving for increasing lags in the partialfollowing autocorrelationformula:<math functiondisplay="block">\phi_{nn} = \frac{\rho(n) - \sum_{k=1}^{n-1} \phi_{n-1, k} \rho(n - k)}{1 - \sum_{k=1}^{n-1} \phi_{n-1, k} \rho(k) }</math>where <math>\phi_{nk} = \phi_{n-1, k} - \phi_{nn} \phi_{n-1,n-k}</math> for <math>1 \leq k \leq n - 1</math>.<ref>{{Cite book |last=Shumway |first=Robert H. |url=http://link.springer.com/10.1007/978-3-319-52452-8 |title=Time Series Analysis and theIts autocorrelationApplications: functionWith R Examples |last2=Stoffer |first2=David S. |date=2017 |publisher=Springer International Publishing |isbn=978-3-319-52451-1 |series=Springer Texts in Statistics |___location=Cham |pages=103-104 |language=en |doi=10.1007/978-3-319-52452-8}}</ref><ref>{{Cite book |last=Enders |first=Walter |url=https://www.worldcat.org/oclc/52387978 |title=Applied econometric time series |date=2004 |publisher=J. Wiley |isbn=0-471-23065-0 |edition=2nd |___location=Hoboken, NJ |pages=65-67 |language=en |oclc=52387978}}</ref>
Partial autocorrelation plots are a commonly used tool for identifying the order of an [[autoregressive model]].<ref name=":0" /> The partial autocorrelation of an AR(''p'') process is zero at lag ''<math>p'' + 1</math> and greater. If the sample autocorrelation plot indicates that an AR model may be appropriate, then the sample partial autocorrelation plot is examined to help identify the order. One looks for the point on the plot where the partial autocorrelations for all higher lags are essentially zero. Placing on the plot an indication of the sampling uncertainty of the sample PACF is helpful for this purpose: this is usually constructed on the basis that the true value of the PACF, at any given positive lag, is zero. This can be formalised as described below.
An approximate test that a given partial correlation is zero (at a 5% [[significance level]]) is given by comparing the sample partial autocorrelations against the critical region with upper and lower limits given by <math>\pm 1.96/\sqrt{n}</math>, where ''n'' is the record length (number of points) of the time-series being analysed. This approximation relies on the assumption that the record length is at least moderately large (say ''<math>n''>30</math>) and that the underlying process has finite second moment.
==References==
{{Reflist}}
*{{cite book |last=Enders |first=Walter |title=Applied Econometric Time Series |___location=New York |publisher=John Wiley |year=2004 |edition=Second |pages=[https://archive.org/details/appliedeconometr00ende_0/page/65 65–67] |isbn=0-471-23065-0 |url-access=registration |url=https://archive.org/details/appliedeconometr00ende_0/page/65}}
{{NIST-PD|http://www.itl.nist.gov/div898/handbook/pmc/section4/pmc4463.htm}}
| 