TOPICS
Search

Ljung-Box Test


The Ljung-Box test is a portmanteau test for whether a group of time-series autocorrelations is zero. For a series of length n, using the first m sample autocorrelations rho^^_k, its statistic is

 Q_(LB)=n(n+2)sum_(k=1)^m(rho^^_k^2)/(n-k).

Under the null hypothesis of no autocorrelation, the statistic is approximately chi-squared distributed. When the statistic is computed from fitted-model residuals, the degrees of freedom are commonly reduced by the number of estimated autoregressive and moving-average parameters. The statistic is a finite-sample modification of the Box-Pierce test.


See also

Autocorrelation, Box-Pierce Test, Durbin-Watson Statistic, Portmanteau Test

Explore with Wolfram|Alpha

References

Ljung, G. M. and Box, G. E. P. "On a Measure of Lack of Fit in Time Series Models." Biometrika 65, 297-303, 1978. https://doi.org/10.1093/biomet/65.2.297.

Cite this as:

Weisstein, Eric W. "Ljung-Box Test." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/Ljung-BoxTest.html

Subject classifications