sdPercentiles function

Test the significance of signal from rolling correlation analysis

Implements the approach of Gershunov et al. (2001) to test the significance of signal from rolling correlation analysis.

sdPercentiles(n = 150, cor = 0.5, width = 5, N = 500, 
              percentiles = c(.05, .95))

Arguments

• n: The length of the series in the rolling correlation analysis.
• cor: The magnitude of raw correaltion betweeen two time series in the rolling correlation analysis.
• width: Window length of the rolling correlation analysis.
• N: Number of Monte Carlo replications for simulations.
• percentiles: Percentiles to be reported for the Monte Carlo distribution of standard deviations of rolling correlations for the given window width.

Details

N samples of correlated white noise series are generated with a magnitude of cor; rolling correlations analysis is applied with the window length of width; Monte Carlo distribution of standard deviations of rolling correlations are generated; and desired percentiles of the MC distribution of standard deviations are reported (Gershunov et al. 2001).

Returns

• rollCorSd.limits: Percentiles of MC distribution of standard deviations of rolling correlations as the test limits.

Author(s)

Haydar Demirhan

Maintainer: Haydar Demirhan haydar.demirhan@rmit.edu.au

References

Gershunov, A., Scheider, N., Barnett, T. (2001). Low-Frequency Modulation of the ENSO-Indian Monsoon Rainfall Relationship: Signal or Noise? Journal of Climate, 14, 2486 - 2492.

Examples

# sdPercentiles(n = 50, cor = 0.5, width = 5, N = 50, 
#              percentiles = c(.025, .975))