Sunday, October 13, 2013

Comparison Of Higher Order Fits For The Monthly Global Land Anomaly

  The "residuals" for the second harmonic fit of the global land anomaly showed what look like a third harmonic so I did two higher order fits. The first assumed there was a third harmonic present in the data and the second was just a polynomial fit. The resulting residuals showed an improvement along the smoothed anomaly for the first fit but the second still showed what looked like the presence of a third harmonic.

The higher order fits seem to favor periodic terms over non-periodic terms. Both projections are roughly in agreement for a period of about 10 years. Each fit had 24 months outside the 3σ limits for a total of 1604 months which is 1.50%. The expected value for the fitted probability distribution is 1.15%.

Supplemental (Oct 14): The statistics above were based on deviations of the anomaly from the unsmoothed 20-year average. I checked the statistics for deviations from the smoothed 20-year average and the results were slightly different. The standard deviation was σ = 0.315. The fit for the 3 harmonics had 19/1604 months (1.185%) outside the 3σ limits. The polynomial fit had 23/1604 months (1.434%) outside the 3σ limits. For a normal distribution with σ = 0.315 one would expect the number of outliers to vary from 13.5 to 21.9 with the expected number being 17.7 (1.103%). The polynomial fit may already be encountering some problems. The terms with the higher powers have more influence away from the center and affect the long term projections more.

No comments: